WO2004084722A1 - Method and apparatus for for identifying features in an ecg signal - Google Patents

Method and apparatus for for identifying features in an ecg signal Download PDF

Info

Publication number
WO2004084722A1
WO2004084722A1 PCT/GB2004/001187 GB2004001187W WO2004084722A1 WO 2004084722 A1 WO2004084722 A1 WO 2004084722A1 GB 2004001187 W GB2004001187 W GB 2004001187W WO 2004084722 A1 WO2004084722 A1 WO 2004084722A1
Authority
WO
WIPO (PCT)
Prior art keywords
peak
time
potential
maximum
value
Prior art date
Application number
PCT/GB2004/001187
Other languages
French (fr)
Inventor
Malcolm Ellis
Rostislav Vychodil
Jiri Pumprla
Original Assignee
Advanced Medical Diagnostics Group Limited
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Advanced Medical Diagnostics Group Limited filed Critical Advanced Medical Diagnostics Group Limited
Publication of WO2004084722A1 publication Critical patent/WO2004084722A1/en

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7239—Details of waveform analysis using differentiation including higher order derivatives
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A61B5/7207—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal of noise induced by motion artifacts

Definitions

  • the present invention relates to a method and apparatus for identifying an R peak in the QRS complex of an ECG signal.
  • Cardiac monitors have been used for some time to measure electrocardiographs of a subject to determine whether their heart is in good physical condition.
  • One set of parameters which is often looked at is the PQRST complex which occurs during each successive heartbeat.
  • the temporal separation between the R wave peak associated with consecutive heartbeats is of particular interest to medical practitioners.
  • the rate at which the R wave peak occurs and variations in the amplitude of those peaks give an indication as to the strength and health of the heart.
  • a well-known method of identifying the occurrence of the R wave peak in a heartbeat is differentiation.
  • a differentiating circuit is used to determine the first derivative of the QRS component of the PQRST complex.
  • the shape of the QRS component is such that the R wave peak is generally a local maximum which results in the first derivative of the R wave peak being zero. If the signal is noisy due to the subject moving around or simply because of artefacts in the recorded ECG another feature may be the local maximum and thus mistakenly identified as an R wave peak. This results in the R-R intervals between successive heartbeats being calculated incorrectly and hence the data being of little use to a medical practitioner.
  • a method for identifying an R peak in the QRS complex of an ECG signal comprising the steps of: identifying a maximum and a minimum in the first derivative of the ECG signal; determining an amplititude difference between the maximum and minimum, identifying a first time corresponding to the maximum; identifying a second time corresponding to a point between the maximum and the minimum at which the first derivative of the ECG signal is substantially zero; determining a temporal difference between the first time and the second time; and identifying the point as a potential R peak if the amplitude difference is greater than a predetermined amplitude value and the temporal difference is greater than a predetermined time value.
  • the method defined above enables an R peak to be identified more reliably and with greater accuracy than existing methods.
  • Calculating a time difference, which takes the shape of the complex into consideration, as well as an amplitude difference enables the temporal occurrence of an R peak, relative to other features in the QRS complex, to be identified quickly and accurately.
  • Figure 4A shows the characteristic wave peaks and troughs of a PQRST complex (labelled P, Q, R, S, and T) which are present in a raw ECG signal.
  • Figures 4B and 5 illustrate the result of differentiating the raw ECG signal to determine the occurrence of the R peak for a given PQRST complex.
  • Q', R', and S' represent the turning points or first moments associated with the Q, R, and S peaks respectively.
  • Figures 4B and 5 show only the QRS portion of the PQRST complex as the P and T waves are suppressed by a differentiating circuit (details of which will follow later) which operates on the ECG signal to determine the first derivative or first moment.
  • R peak identification generally involves simply comparing the amplitude of a raw ECG signal against a fixed or a varying threshold value. If the amplitude is greater than or equal to the fixed value or the varied value a peak is accepted as an R peak. However, if a monitor is incorrectly positioned or if the subject is moving around the R peak may not necessarily be the feature in a given PQRST complex which has the greatest amplitude. In particular, if a subject is exercising the signal may be very noisy and the amplitude of successive R peaks may vary quite considerably.
  • the method and apparatus of the present invention utilise the following points of interest in a differentiated ECG signal as illustrated in Figure 5 :
  • the difference is calculated between the maximum amplitude and the minimum amplitude and the result compared with a predetermined amplitude value.
  • the temporal occurrence of the maximum (T max ) is measured relative to a predetermined temporal reference point.
  • a temporal difference is also calculated between the time of occurrence of the turning point (T down ) an & tne maximum (T raax ) and the result is compared to a predetermined time value. If both the amplitude difference and the temporal difference are greater than or equal to the predetermined amplitude value and predetermined time value respectively the temporal occurrence of the turning point is identified as that of a potential R peak.
  • the maximum is a global maximum.
  • the minimum is a global minimum.
  • the first time is measured relative to a predetermined temporal reference point.
  • the predetermined temporal reference point may be a global maximum or a global minimum of the differentiated signal, or a turning point of the ECG signal.
  • the predetermined temporal reference point is the last recorded valid R peak.
  • the predetermined temporal reference point is the zero state of a free-running 16-bit timer. The zero state being the overflow condition of the timer.
  • the predetermined amplitude value is directly dependent upon the amplitude of a previously recorded beat amplitude in the ECG signal.
  • the predetermined amplitude value is between 5% and 100% of an amplitude range of an A/D converter.
  • the predetermined amplitude value is between 50% and 70% of an amplitude range of an A/D converter.
  • the predetermined time value is between 5msec and 10 msec. Most preferably, predete ⁇ riined time value is approximately 8 msec.
  • the method further comprises the steps of setting a time period after the identification of the potential R peak and rejecting the potential peak if any further fiducial point occurs within the period. This prevents fiducial points which occur too close together temporally being accepted as valid points.
  • the method further comprises the steps of setting a time period after the identification of the potential R peak and accepting the potential R peak as a valid R peak if no fiducial point occurs within the period. This prevents fiducial points which occur too close together temporally being accepted as valid points.
  • the method further comprises the step of calculating the time interval between the valid R peak and the last recorded R peak to evaluate an R-R interval.
  • the method may include the step of setting a time period substantially equal to 100 msec. In other embodiments, the time period may be between 50msec and 200msec.
  • the method further comprises the step of setting the current calculated amplitude difference to be the next predetermined amplitude value.
  • the method further comprises the step of decrementing the predetermined amplitude value based on a previous value.
  • the method includes the step of decrementing the predete ⁇ riined amplitude value using a dynamical threshold.
  • the dynamic threshold may take the form of a decaying exponential. Use of such a decaying threshold amplitude results in the R peak being identified more reliably and accurately.
  • the lowest threshold value of the predetermined amplitude value occurs in a time interval where the probability of an R peak occurring is highest.
  • the probability of finding an R peak when the lowest threshold value of the predetermined amplitude value occurs is greater than 80%. More preferably, the probability of finding an R peak when the lowest threshold value of the predetermined amplitude value occurs is greater than 90%.
  • the amplitude threshold decay is delayed for a predetermined time period after the identification of an R peak. Incorporating a time delay reduces the probability of a false R peak being detecting in this time period.
  • the step of evaluating the first derivative of the signal is carried out using hardware.
  • the step of identifying R peaks is carried out using a microprocessor.
  • a data processing apparatus for identifying an R peak in the QRS complex of an ECG signal, the apparatus comprising: a microprocessor; a memory in communication with the microprocessor and storing instructions executable by the microprocessor to: identify a maximum and a minimum in the amplitude of the first derivative of the ECG signal; determine an amplitude difference between the maximum and minimum; identify a first time corresponding to the maximum; identify a second time corresponding to a point between the maximum and the minimum at which the first derivative of the ECG signal is substantially zero; determine a temporal difference between the first time and the second time; and identify the point as a potential R peak if the amplitude difference is greater than a predetermined amplitude value and the temporal difference is greater than a predetermined time value.
  • Theapparatus defined above enables an R peak to be identified more reliably and with greater accuracy than existing methods.
  • Calculating a time difference, which takes the shape of the complex into consideration, as well as an amplitude difference enables the temporal occurrence of an R peak, relative to other features in the QRS complex, to be identified quickly and accurately.
  • the microprocessor sets a time period after the identification of the potential R peak and rejects the potential R peak if any farther potential R peak occurs within the period.
  • the microprocessor sets a time period after the identification of the potential R peak and accepts the potential R peak as a valid R peak if no further potential R peak occurs within the period.
  • the microprocessor calculates the time interval between the valid R peak and the last recorded R peak to evaluate an R-R interval.
  • the microprocessor sets a time period of approximately 100 ms.
  • the microprocessor sets the current calculated amplitude difference to be the next predetermined amplitude value.
  • the microprocessor decrements the predetermined amplitude value based on a previous value.
  • the first derivative of the ECG signal is evaluated out using hardware
  • R peak identification (especially in noisy signal) is differentiation. This stems from a mathematical theorem regarding the identification of local extremes. In existing systems, differentiation may be performed digitally after rough ECG signal sampling. However, there are some limitations in the practical implememtation of this. If the ECG signal is contaminated (e.g during physical exercise) by isoline drifting an A/D converter with a high dynamic range will be required. In the present method and apparatus, the first part of the peak identification process (differentiation) is conducted using hardware, the first signal filter, to produce a derived signal. The second part of the R peak identification is carried out under software control in the micro-controller. This enables a low dynamic range 8-bit A/D converter to be used which reduces the overall costs involved.
  • Figure 1 is a schematic illustrating how a transmitter of an ECG monitor in accordance with the present invention operates
  • Figure 2 is a schematic illustrating how a receiver associated the transmitter in Figure 1 operates;
  • Figure 3 shows a circuit diagram of the transmitter of Figure 1;
  • Figure 4A shows an amplified PQRST complex for a single heartbeat as measured in an ECG
  • Figure 4B shows the result of differentiating the signal in Figure 4A
  • Figure 5 shows a more detailed view of the differentiated signal of Figure 4B
  • Figure 6 is a flow diagram illustrating the R-peak identification method.
  • Figure 1 shows a transmitter 10 which can be placed on a subj ects body to record a cardiac signal .
  • the transmitter 10 maybe attached to a belt (not shown) and worn around the subjects chest.
  • the transmitter 10 has two electrodes 12 which can be attached to the subjects body to record an electrocardiograph or ECG.
  • the signal is dealt with by a processor (details of which will follow) which includes first and second signal filters 14, 16 and a micro-controller 18.
  • the recorded signal passes to the first and second signal filters 14, 16 which separate the signal into first and second signal components.
  • the first signal component is then manipulated by the processor to evaluate the temporal spacing between consecutive R wave peaks in the QRS complex of the electrocardiograph signal.
  • the first and second signal components are then fed into the micro-controller 18 where the sampling rate of the second component, which effectively consists of the raw ECG, is reduced.
  • the micro-controller 18 feeds the first signal component and reduced sampling rate second signal component into a digitised data stream and on to a radio transmitting module 20 for telemetric transmission from a transmission antenna 22.
  • ECG signal is recorded by means of apair of electrodes 12 which can be placed in contact with the skin of a subject.
  • the recorded ECG signal initially passes through a pair of low-pass RC filter circuits 100, 102 which consist of a resistor R3 and a capacitor CI, and a resistor R4 and a capacitor C2, respectively.
  • resistors R3 and R4 are set to 10 K ohm and capacitors CI and C2 are set to 100 pF.
  • the outputs from the filter circuits 100, 102 form the inputs to a differential amplifier 104 based on an AD 627 integrated circuit.
  • the low-pass filter circuits 100, 102 filter high frequency signals which result from radio module operation and protect the amplifier inputs against damage by high level impulse over- voltage (electrostatic discharge voltage).
  • the differential amplifier 104 differentiates the signal from background noise to create a relatively "clean" signal.
  • resistors Rl and R2 serve as a DC bias supply which is necessary for proper functioning of the differential amplifier 104.
  • the values of the resistors Rl and R2 (101, 103) are selected to provide a high impedance to the differential amplifier inputs.
  • resistors Rl and R2 are set to 8 M ohm.
  • a DC bias is created by a resistor divider R21, R26 (105) and blocked by a capacitor C14.
  • the gain of the differential amplifier 104 is set by a further resistor R5 (106).
  • resistors R21, R26 and R5 are respectively set to 4 K ohm, 4 K ohm, and 12 K ohm and capacitor C 14 set to 220 nF.
  • the gain of the differential amplifier 104 is selected to prevent saturation from any ECG signal baseline drift.
  • the rough ECG signal 107 is divided into two signal components by the first and second signal filters 14, 16 which adjust the signal spectra and provide sufficient amplification necessary for the channels of an A/D converter which is integrated on the same silicon chip as the micro-controller 18.
  • the first and second signal components are fed into a pair of high-pass filter circuits 108, 111.
  • the first of these high-pass filter circuits 108 includes a resistor R12 and a capacitor C5 the values of which are set such that the circuit 108 differentiates the signal.
  • the second high-pass filter circuit 111 also has a resistor Rl 1 and a capacitor C4 but does not differentiate the signal.
  • resistors R12 and Rl 1 are set to 100 K ohm and 1 M ohm respectively, and capacitors C5 and C4 set to 47 nF and 220 nF respectively.
  • the frequency spectrum of the QRS complex contains frequencies in the range of around 20 to 100 Hz.
  • Undesirable signals such as cardiac artefacts have a 1/f-type spectrum with their main components being located below 20 Hz. It is therefore useful to process the rough ECG signal using a high-pass filter 108 before A/D conversion. In practice a cut-off frequency of around 34 Hz is used.
  • Filtering and amplifying of the first and second signal components 109, 112 of the ECG signal are performed by OP 295 single-supply micro-power operating amplifiers 110, 113 with "rail- to-rail" capability.
  • These operational amplifiers 110, 113 allow the A/D converter reference voltage to be used as an operating amplifier power supply with no loss in A/D converter input range.
  • Both signal, branches 114, 116 act as an low-pass filters, the cut-off frequencies determined by resistor Rl 0 and capacitor C3, and resistor Rl 5 and capacitor C6, respectively.
  • resistors R10 and Rl 5 are set to 100 K ohm and capacitors C3 and C6 are set to 2.2 nF.
  • Resistors R8 and R13 determine the first and second channel gain for the signal branches.
  • Resistors R9 and R14 are used for the differential amplifier 104 input bias current compensation and to avoid unwanted DC level shift in output signal.
  • resistors R8, R13, R9 and R14 are respectively set to 3 K ohm, 1 K ohm, 1 M ohm, and 100 K ohm.
  • the overall operations of the transmitter 10 are controlled by a micro-controller 18.
  • a PIC16C711 micro-controller is used (manufactured by Microchip). This is a low power device is based on "Dual Bus Harvard RISC" architecture and provides high computational power relative to its power consumption.
  • the micro-controller 18 has four integrated A/D converter channels, only three of which are used, which allow direct connection to the first and second signal filters 14, 16 and thus significantly reduce the complexity of the transmitter circuitry.
  • the main functions performed by the micro-controller 18 are sampling and A/D conversion of the ECG signal from both signal branches.
  • the sampling rate for the first signal component 109 of the ECG signal is around 1000 Hz. This corresponds to a time interval of around 1ms which is necessary to obtained accurate R-R intervals from the ECG signal.
  • the sampling rate for the second component 112 may be set at somewhere between 100 and 500 Hz.
  • Further functions performed by the micro-controller 18 include coding of measured R-R interval values (first signal component), ECG signal samples (second signal component) and battery status into a serial data stream using a dedicated transfer protocol prior to telemetric transfer, charging control of internal lithium-ion rechargeable battery (two-stage CC and CV charging method), diagnostics and service functions.
  • the recorded ECG signal is amplified and digitally processed directly in the transmitter 10 (on the patient's body) and an instantaneous heart rate (R-R intervals) precisely derived.
  • This information together with the raw ECG signal and battery status information are encoded by a transfer protocol into a serial code (suitable for optimal narrowband radio channel exploitation) and transmitted in the UHF band by the radio transmitting module 20 and transmission ariel 22.
  • the transmitted signal is in digital form.
  • the present invention incorporates a technique known as diversity reception which eliminates fading of a transmitted signal. Fading is caused by multipath wave propagation, which results from reflection and absorption of transmitted radio waves between a transmitter and a receiver. Multiple waves received by an ariel will be added together according to the principle of vector summation. This means that in certain conditions (when the waves are in opposite phase) the vector sum may be close to zero and transferred information may be lost.
  • An effective way of overcoming this problem is to use two spaced apart antennae 32, 34 which increases the probability of a good signal level being received in one antenna if the signal received in the other antenna suffers interference. Summing signals from the two antennae is likely to encounter the problem of out of phase signals cancelling each other out.
  • the signals received by the two receiver antennae 32, 34 are processed by two separate receiving modules 36, 38 and are then switched by the analogue switching circuit which determines the best quality signal.
  • the selected signal is processed in a data recovery circuit (slicer) which separates the digital data from analog signal.
  • the transmitted signal is received by one or both of a pair of receiver antennae 32, 34 which form part of a remote receiver 30.
  • the receiver antennae 32, 34 operate under diversity reception to ensure that the best quality signal available is received by the receiver 30.
  • he received signal (s) pass through a pair of receiver module 36, 38 which pass the signal(s) to a receiver micro-controller 44 either directly or via an analogue switch 40 and a data recovery unit 42.
  • Data processed by the micro-controller 44 by a suitable connector 46 to and from a PC.
  • the connector 46 is an RS232-type interface.
  • R wave peak identification can become complicated as local maxima/minima occur for Q and S wave peaks during each heartbeat as shown in Figure 4.
  • Noise peaks and troughs associated with ECG signals can also lead to further maxima/minima. This is particularly true of very noisy ECG signals which are obtained when the subject is moving or exercising.
  • test conditions are set up according to characteristic Rpeak properties to provide reliable identification of R wave peaks even in a noisy ECG signal.
  • FIG. 4 The wave peaks and troughs of the PQRST complex are labelled P, Q, R, S, and T.
  • Figures 4 and 5 show respectively the raw ECG signal before and after differentiation.
  • Q', R 1 , and S' are the turning points or first moments associated with the Q, R, and S peaks respectively.
  • Figure 5 shows only the QRS portion of the PQRST complex as the P and T waves are suppressed by the first high-pass filter 108, the differentiating circuit .
  • the time and voltage scales have been magnified for clarity. The y-ordinate which does not represent millivolts but a numeric value resulting from A/D conversion. A nominal zero level of 128 has been adopted.
  • the method of the present invention is more flexible than existing methods as it takes into consideration the fact the shape of the QRS complex can vary considerably which results in variations in the amplitude of the Rpeak. This can be the case if the electrodes are incorrectly positioned on a subject. This often results in the R and/or S peaks increasing in amplitude.
  • the dependence of the pre-processed signal amplitude on the whole QRS complex negates the effect of ECG electrode positioning as the amplitude difference between the global maximum and the global minimum in the differentiated signal is practically constant for any ECG signal shape.
  • the occurrence of points 3) and 4) is stored in a temporary memory. These points do not relate to local maximum or minimum but global extremes computed on a time interval defined by the turning points associated with Q, R, and S peaks and troughs.
  • the global maximunVminimum searching algorithm of the present invention always starts in response to the occurrence of a first turning point i.e. T up . This eliminates false identification of local extreme, because multiple local extremes can appear on the QRS wave as a result of signal noise.
  • the zero of the y-ordinate in Figure 5 relates to a zero signal level or isoelectrical line of the A/D converter i.e. thel28 value corresponds to the zero signal level.
  • each downward zero crossing point could potentially be an R peak occurence.
  • these times are written to a memory as a preliminary R peak occurrence or preliminary fiducial point.
  • the present method uses further computational checks to verify each preliminary R peak occurrence as a true R peak.
  • the whole QRS complex for that heartbeat needs to be considered.
  • the R wave peak is confirmed only after the signal minimum 4) or MTN has been identified.
  • the algorithm used in the present invention tests the peak-peak amplitude of each sinusoidal-like waveform (MAX - MTN) to determine whether it exceeds a predetermined limit (AMP). If the amplitude is greater than the predetermined limit the algorithm moves to a further test condition. If the amplitude condition fails, the potential R peak or fiducial point is rejected and the algorithm continues searching for the true R peak
  • the next step of the algorithm considers the time interval between 3) the signal maximum or T max and 2) the downward zero crossing T down .
  • This interval, T down -T max is directly related to the shape of the QRS complex. If the calculated time difference is less than a predetermined value Q lim , a false R peak or fiducial point is registered. Such false points are usually caused by myopotential noise. If the time condition test fails, the preliminary fiducial point is rejected and the algorithm continues searching for a true or valid R peak.
  • the potential fiducial point is passed to another part of the memory for storage as a fiducial point.
  • one further check is conducted to ensure that an a valid R wave peak has been identified.
  • This additional checking condition is invoked to prevent R-R intervals shorter than a predetermined minimum value (M3NRR) (usually around 100 msec) being recorded.
  • M3NRR predetermined minimum value
  • R-R intervals shorter than 100 msec would result in data processing problems. For these reasons, supposed R wave peaks occurring within such a short time interval after the last positive identification are rejected.
  • next identified peak takes longer than the set time period (MTNRR) to appear it is accepted as a fiducial point.
  • MNRR set time period
  • the R-R interval value is computed as a difference between current R peak or fiducial point time and the previous fiducial point time (stored in memory from last cycle). The computed value is stored and then passed on for transmission purposes.
  • the test conditions consist of comparing the sine-like wave peak-peak . amplitude (MAX - MEN) with a predetermined value.
  • amplitude MAX - MEN
  • the ECG amplitude changes from beat to beat, which means that the amplitude of the R wave peak may vary continuously. This is particularly true if a subject is moving while an ECG is being recorded.
  • use of a fixed value or threshold for comparing R wave peaks can be unreliable for positive fiducial point identification.
  • a low threshold can result in false positive beat identification (noise will be interpreted as heartbeat) and a high threshold value may result in false negative identification error (heartbeats missing).
  • Methods based on solely on comparing the amplitude against a set threshold have proven to be unreliable, often resulting in identification of false R wave peaks, lost R wave peaks and a lack of precision in the R-R intervals evaluated.
  • the present invention uses a dynamic or adaptive threshold which changes periodically.
  • the amplitude of the last successfully identified wave is designated as the threshold.
  • This threshold value is periodically decremented at set time intervals ⁇ T.
  • the values of the dynamic threshold are given by following geometric progression:
  • the initial value U 0 represents the previous wave amplitude and it can be deduced that the equivalent time constant t is given by following expression:
  • ⁇ time constant of exponentially falling curve
  • ⁇ T time period of the decrementing algorithm (period of iteration)
  • q progression quotient which can be expressed more usefully by the following equation:
  • D is a dividing factor incorporated into the software of the micro-controller 18, and which is set in the source code before compilation.
  • the micro-controller software divides the last stored threshold value by the dividing factor D and the new threshold value is obtained by subtraction.
  • the dividing factor D value is restricted to 2 n numbers only due to limited mathematical capabilities of the micro-controller.
  • the exponentially falling dynamic threshold prevents false beat identification as the probability of beat occurrence increases in the time, counted from the last heartbeat occurrence moment, corresponding to the falling thresholds.
  • a high level of dynamic threshold just after the heartbeat prevents disturbing noise pulses being falsely identified.
  • R-R interval measuring algorithm parameters dynamic threshold behaviour, time restricting condition etc.
  • FLO DIAGRAM DESCRIPTION All of the R-R interval measuring algorithm parameters (dynamic threshold behaviour, time restricting condition etc.) can be selected before source code compilation to optimise the algorithm performance.
  • the micro-controller incorporates a 16 bit free-running timer, having a timer clock fremjencv is 1000 Hz.
  • the timer is incremented once every millisecond, thus the clock rate 1000 Hz and 16 bit capacity of the timer means that it overflows roughly every 64 seconds.
  • This timescale is much longer than any time interval between features which appear in the ECG signal. Therefore no absolute timescale is required with all relevant time points being identified and stored relative to one another rather than relative to a set start or end point. In particular, the time intervals between points of interest are calculated by relative time point subtractions. There is no problem even when measured interval is falling to timer overflow moment because of binary mathematics properties (MODULO 2 16 computing). Mathematical operations in 16-bit registers automatically perform MODULO operation without any additional programming. The instant timer value is labelled as T in the flow diagram of Figure 6.
  • the main algorithm loop always begins using a signal which has been differentiated and has then undergone A/D conversion.
  • the algorithm hunts or searches for all of the important points in the differentiated QRS waveform i.e. global maximum (MAX), global minimum (MTN), a downward zero crossing (T down ), and the temporal position of the global maximum (T max ).
  • the main loop also contains tests which determine if particular actions need. to be performed at particular times i.e. sending R-R time interval values for telemetric transmission or controlling the dynamical threshold decay.
  • the global maximum searching algorithm permanently compares the current sample value SAMPLE(n) to a value MAX stored in a memory location. If the current sample is greater than the stored MAX value then the stored value is replaced by the current sample (i.e. MAX is updated). The initial value given to the value MAX is reset before each maximum searching cycle.
  • the nominal zero value is first encountered at T up (first upward zero crossing). This can be seen in Figure 5. It also useful to note that the global maximum MAX must be located on the curve segment between T up ( first upward zero crossing) and T down (the downward zero crossing) due to the nature and shape of the differentiated QRS complex. As the searching cycle progresses the algorithm is incremented and each successive temporary maximum is stored (both its value MAX and time of occurrence Tmax) until the global maximum value is reached. The maximum values are recorded between points 1) and 3) on Figure 5.
  • the search cycle for the minimum begins after value T down is reached and the search for the global minimum is performed in a manner similar to that described above.
  • the memory location MIN is updated every time the current sample SAMPLE(n) is less than a stored value
  • the minimum searching algorithm is reset (nominal zero of 128 is stored to memory location MTN) when T do ⁇ , n (downward zero crossing point) is encountered.
  • the difference compared to the maximum searching procedure is that the algorithm is not waiting until a new upward zero crossing is reached and a definitive or global mir ⁇ num is found. Instead, the algorithm checks every temporary minimum to see whether that point meets the amplitude condition (MAX - MTN > AMP). When the amplitude condition is met then time condition (Tdown - Tmax > Qlim) is checked and if both are satisfied (i.e. a beat has been detected) then following actions are performed in the micro-controller:
  • the downward zero crossing is identified when the current sample SAMPLE(n) is zero or drops below value 128 (zero voltage level). Applying only this condition is not sufficient as all samples below the nominal zero level of 128 will be considered to be downward zero crossing. To overcome this an additional condition is used, namely using the last sample SAMPLE(n-l).
  • the downward zero crossing point is used as a preliminary fiducial point and also serves for minimum searching algorithm reset as described above.
  • the downward zero crossing point 5 represents a local maximum and therefore is a potential fiducial point or R peak.
  • the upward zero crossing point is detected in a similar manner (current sample is zero or above zero line and the last sample is below zero line) and it is used for maximum search algorithm reset purpose only.
  • the R-R interval value is finally calculated and sent for transmission if no new fiducial point is detected during a time period MINRR. If such a new fiducial point is detected during the MINRR time period a previously saved fiducial point will be replaced by this new one and new time period MTNRR will start. This methodology is ensured by the following "send time” test:
  • - R-R value is calculated as a difference between new and old (saved from last beat ientification) fiducial points
  • the current sample SAMPLE(n) is saved at the end of whole loop as SAMPLE(n-l) for next cycle usage.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Cardiology (AREA)
  • Surgery (AREA)
  • Medical Informatics (AREA)
  • Veterinary Medicine (AREA)
  • Physics & Mathematics (AREA)
  • Public Health (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Psychiatry (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

A method for identifying an R peak in the QRS complex of an ECG signal includes the steps of identifying a maximum (MAX,3) and a minimum (MIN,4) in the first derivative of the ECG signal and determining an amplititude difference between the maximum and minimum. A first time (Tmax) corresponding to the maximum (MAX) is identified as is a second time (Tdown) corresponding to a point between the maximum (MAX,3) and the minimum (MIN,4) at which the first derivative of the ECG signal is substantially zero. A temporal difference between the first time (Tmax) and the second time (Tdown) is determined. The point is identified as a potential R peak if the amplitude difference is greater than a predetermined amplitude value (AMP) and the temporal difference is greater than a predetermined time value (Qlim).

Description

Method and apparatus for identifying features in an ECG signal
The present invention relates to a method and apparatus for identifying an R peak in the QRS complex of an ECG signal.
Cardiac monitors have been used for some time to measure electrocardiographs of a subject to determine whether their heart is in good physical condition. One set of parameters which is often looked at is the PQRST complex which occurs during each successive heartbeat. The temporal separation between the R wave peak associated with consecutive heartbeats is of particular interest to medical practitioners. The rate at which the R wave peak occurs and variations in the amplitude of those peaks give an indication as to the strength and health of the heart.
A well-known method of identifying the occurrence of the R wave peak in a heartbeat is differentiation. A differentiating circuit is used to determine the first derivative of the QRS component of the PQRST complex. The shape of the QRS component is such that the R wave peak is generally a local maximum which results in the first derivative of the R wave peak being zero. If the signal is noisy due to the subject moving around or simply because of artefacts in the recorded ECG another feature may be the local maximum and thus mistakenly identified as an R wave peak. This results in the R-R intervals between successive heartbeats being calculated incorrectly and hence the data being of little use to a medical practitioner.
Other methods involve comparison of the R wave amplitude for each heartbeat against a fixed threshold. These methods can also lead to false identification of an R wave peak as the ECG signal amplitude, and thus R wave peak amplitude, can vary quite considerably from heartbeat to heartbeat. This is particularly true if a subject is moving around or exercising.
To alleviate this problem adaptive thresholds have been used in which the amplitude of a correctly identified R wave peak is used as the basis for determining whether a feature in the ECG is indeed the next R wave peak. However, due to the noise and artefacts in the recorded
ECG, the use of an adaptive threshold can still result in incorrectly identified R wave peaks. According to a first aspect of the present invention there is provided a method for identifying an R peak in the QRS complex of an ECG signal, the method comprising the steps of: identifying a maximum and a minimum in the first derivative of the ECG signal; determining an amplititude difference between the maximum and minimum, identifying a first time corresponding to the maximum; identifying a second time corresponding to a point between the maximum and the minimum at which the first derivative of the ECG signal is substantially zero; determining a temporal difference between the first time and the second time; and identifying the point as a potential R peak if the amplitude difference is greater than a predetermined amplitude value and the temporal difference is greater than a predetermined time value.
The method defined above enables an R peak to be identified more reliably and with greater accuracy than existing methods. Calculating a time difference, which takes the shape of the complex into consideration, as well as an amplitude difference enables the temporal occurrence of an R peak, relative to other features in the QRS complex, to be identified quickly and accurately.
Figure 4A shows the characteristic wave peaks and troughs of a PQRST complex (labelled P, Q, R, S, and T) which are present in a raw ECG signal. Figures 4B and 5 illustrate the result of differentiating the raw ECG signal to determine the occurrence of the R peak for a given PQRST complex. In Figures 4B and 5, Q', R', and S' represent the turning points or first moments associated with the Q, R, and S peaks respectively. Figures 4B and 5 show only the QRS portion of the PQRST complex as the P and T waves are suppressed by a differentiating circuit (details of which will follow later) which operates on the ECG signal to determine the first derivative or first moment.
In Figure 5 the time and voltage scales have been magnified for clarity. The y-ordinate which does not represent millivolts but a numeric value resulting from A D conversion (details of which follow later). A nominal zero level of 128 has been adopted with values between 0 and 127 (inclusive) designated as negative and those between 129 and 255 (inclusive) designated as positive. The signal shown in Figure 4B no longer represents genuine Q, R and S waves due to the effects of a first signal filter (details of which follow later). However, this change of shape does not ultimately affect the measurement of R-R time intervals.
The important features which can be obtained from the differentiated signal shown in Figures 4B and 5 are:
1) the downward zero crossing corresponding to the R wave peak which can be used as a precise fiducial point (shown as R'or Tdown); and 2) the amplitude of the waveform which is directly proportional to that of the original QRS complex, thus enabling it to be used for accurate Rpeak identification. •
Known methods of R peak identification generally involve simply comparing the amplitude of a raw ECG signal against a fixed or a varying threshold value. If the amplitude is greater than or equal to the fixed value or the varied value a peak is accepted as an R peak. However, if a monitor is incorrectly positioned or if the subject is moving around the R peak may not necessarily be the feature in a given PQRST complex which has the greatest amplitude. In particular, if a subject is exercising the signal may be very noisy and the amplitude of successive R peaks may vary quite considerably.
The method and apparatus of the present invention utilise the following points of interest in a differentiated ECG signal as illustrated in Figure 5 :
1)- the temporal position of the maximum amplitude, associated with the maximal positive ECG slope (Tmax);
2) the temporal position of a turning point associated with an R peak (Tdown);
3) the maximum amplitude (MAX); and 4) the minimum amplitude, associated with maximal negative ECG slope (MTN).
The difference is calculated between the maximum amplitude and the minimum amplitude and the result compared with a predetermined amplitude value. The temporal occurrence of the maximum (Tmax) is measured relative to a predetermined temporal reference point. A temporal difference is also calculated between the time of occurrence of the turning point (Tdown) an& tne maximum (Traax) and the result is compared to a predetermined time value. If both the amplitude difference and the temporal difference are greater than or equal to the predetermined amplitude value and predetermined time value respectively the temporal occurrence of the turning point is identified as that of a potential R peak.
Preferably, the maximum is a global maximum. Conveniently, the minimum is a global minimum.
In preferred embodiments, the first time is measured relative to a predetermined temporal reference point. The predetermined temporal reference point may be a global maximum or a global minimum of the differentiated signal, or a turning point of the ECG signal. Preferably, the predetermined temporal reference point is the last recorded valid R peak. Most preferably, the predetermined temporal reference point is the zero state of a free-running 16-bit timer. The zero state being the overflow condition of the timer.
In preferred embodiments, the predetermined amplitude value is directly dependent upon the amplitude of a previously recorded beat amplitude in the ECG signal. Preferably,- the predetermined amplitude value is between 5% and 100% of an amplitude range of an A/D converter. Most preferably, the predetermined amplitude value is between 50% and 70% of an amplitude range of an A/D converter.
Conveniently, the predetermined time value is between 5msec and 10 msec. Most preferably, predeteπriined time value is approximately 8 msec.
Preferably, the method further comprises the steps of setting a time period after the identification of the potential R peak and rejecting the potential peak if any further fiducial point occurs within the period. This prevents fiducial points which occur too close together temporally being accepted as valid points. Conveniently, the method further comprises the steps of setting a time period after the identification of the potential R peak and accepting the potential R peak as a valid R peak if no fiducial point occurs within the period. This prevents fiducial points which occur too close together temporally being accepted as valid points.
In preferred embodiments, the method further comprises the step of calculating the time interval between the valid R peak and the last recorded R peak to evaluate an R-R interval.
The method may include the step of setting a time period substantially equal to 100 msec. In other embodiments, the time period may be between 50msec and 200msec.
Conveniently, the method further comprises the step of setting the current calculated amplitude difference to be the next predetermined amplitude value.
Preferably, the method further comprises the step of decrementing the predetermined amplitude value based on a previous value. Conveniently, the method includes the step of decrementing the predeteπriined amplitude value using a dynamical threshold. The dynamic threshold may take the form of a decaying exponential. Use of such a decaying threshold amplitude results in the R peak being identified more reliably and accurately.
Preferably, the lowest threshold value of the predetermined amplitude value occurs in a time interval where the probability of an R peak occurring is highest. Conveniently, the probability of finding an R peak when the lowest threshold value of the predetermined amplitude value occurs is greater than 80%. More preferably, the probability of finding an R peak when the lowest threshold value of the predetermined amplitude value occurs is greater than 90%.
In preferred embodiments, the amplitude threshold decay is delayed for a predetermined time period after the identification of an R peak. Incorporating a time delay reduces the probability of a false R peak being detecting in this time period.
Preferably, the step of evaluating the first derivative of the signal is carried out using hardware. Conveniently, the step of identifying R peaks is carried out using a microprocessor.
According to a second aspect of the invention there is provided a data processing apparatus for identifying an R peak in the QRS complex of an ECG signal, the apparatus comprising: a microprocessor; a memory in communication with the microprocessor and storing instructions executable by the microprocessor to: identify a maximum and a minimum in the amplitude of the first derivative of the ECG signal; determine an amplitude difference between the maximum and minimum; identify a first time corresponding to the maximum; identify a second time corresponding to a point between the maximum and the minimum at which the first derivative of the ECG signal is substantially zero; determine a temporal difference between the first time and the second time; and identify the point as a potential R peak if the amplitude difference is greater than a predetermined amplitude value and the temporal difference is greater than a predetermined time value.
Theapparatus defined above enables an R peak to be identified more reliably and with greater accuracy than existing methods. Calculating a time difference, which takes the shape of the complex into consideration, as well as an amplitude difference enables the temporal occurrence of an R peak, relative to other features in the QRS complex, to be identified quickly and accurately.
Conveniently, the microprocessor sets a time period after the identification of the potential R peak and rejects the potential R peak if any farther potential R peak occurs within the period. Preferably, the microprocessor sets a time period after the identification of the potential R peak and accepts the potential R peak as a valid R peak if no further potential R peak occurs within the period.
In preferred embodiments, the microprocessor calculates the time interval between the valid R peak and the last recorded R peak to evaluate an R-R interval.
Preferably, the microprocessor sets a time period of approximately 100 ms.
Conveniently, the microprocessor sets the current calculated amplitude difference to be the next predetermined amplitude value.
Preferably, the microprocessor decrements the predetermined amplitude value based on a previous value.
In preferred embodiments, the first derivative of the ECG signal is evaluated out using hardware
The basis of R peak identification (especially in noisy signal) is differentiation. This stems from a mathematical theorem regarding the identification of local extremes. In existing systems, differentiation may be performed digitally after rough ECG signal sampling. However, there are some limitations in the practical implememtation of this. If the ECG signal is contaminated (e.g during physical exercise) by isoline drifting an A/D converter with a high dynamic range will be required. In the present method and apparatus, the first part of the peak identification process (differentiation) is conducted using hardware, the first signal filter, to produce a derived signal. The second part of the R peak identification is carried out under software control in the micro-controller. This enables a low dynamic range 8-bit A/D converter to be used which reduces the overall costs involved.
The present invention will now be described, by way of example only, with reference to the following figures in which: Figure 1 is a schematic illustrating how a transmitter of an ECG monitor in accordance with the present invention operates;
Figure 2 is a schematic illustrating how a receiver associated the transmitter in Figure 1 operates; Figure 3 shows a circuit diagram of the transmitter of Figure 1;
Figure 4A shows an amplified PQRST complex for a single heartbeat as measured in an ECG; Figure 4B shows the result of differentiating the signal in Figure 4A; Figure 5 shows a more detailed view of the differentiated signal of Figure 4B; and Figure 6 is a flow diagram illustrating the R-peak identification method.
The apparatus of the present invention is shown schematically in Figures 1 and 2. Figure 1 shows a transmitter 10 which can be placed on a subj ects body to record a cardiac signal . The transmitter 10 maybe attached to a belt (not shown) and worn around the subjects chest. The transmitter 10 has two electrodes 12 which can be attached to the subjects body to record an electrocardiograph or ECG. The signal is dealt with by a processor (details of which will follow) which includes first and second signal filters 14, 16 and a micro-controller 18.
The recorded signal passes to the first and second signal filters 14, 16 which separate the signal into first and second signal components. The first signal component is then manipulated by the processor to evaluate the temporal spacing between consecutive R wave peaks in the QRS complex of the electrocardiograph signal. The first and second signal components, are then fed into the micro-controller 18 where the sampling rate of the second component, which effectively consists of the raw ECG, is reduced. The micro-controller 18 feeds the first signal component and reduced sampling rate second signal component into a digitised data stream and on to a radio transmitting module 20 for telemetric transmission from a transmission antenna 22.
The operation of the transmitter will now be described in greater detail with reference to Figure 3.An ECG signal is recorded by means of apair of electrodes 12 which can be placed in contact with the skin of a subject. The recorded ECG signal initially passes through a pair of low-pass RC filter circuits 100, 102 which consist of a resistor R3 and a capacitor CI, and a resistor R4 and a capacitor C2, respectively. In the present embodiment, resistors R3 and R4 are set to 10 K ohm and capacitors CI and C2 are set to 100 pF. The outputs from the filter circuits 100, 102 form the inputs to a differential amplifier 104 based on an AD 627 integrated circuit. The low-pass filter circuits 100, 102 filter high frequency signals which result from radio module operation and protect the amplifier inputs against damage by high level impulse over- voltage (electrostatic discharge voltage).The differential amplifier 104 differentiates the signal from background noise to create a relatively "clean" signal.
Two resistors Rl and R2 (101, 103) serve as a DC bias supply which is necessary for proper functioning of the differential amplifier 104. The values of the resistors Rl and R2 (101, 103) are selected to provide a high impedance to the differential amplifier inputs. In the present embodiment, resistors Rl and R2 are set to 8 M ohm. A DC bias is created by a resistor divider R21, R26 (105) and blocked by a capacitor C14. The gain of the differential amplifier 104 is set by a further resistor R5 (106). In the present embodiment, resistors R21, R26 and R5 are respectively set to 4 K ohm, 4 K ohm, and 12 K ohm and capacitor C 14 set to 220 nF. The gain of the differential amplifier 104 is selected to prevent saturation from any ECG signal baseline drift.
After basic amplification by the differential amplifier 104 the rough ECG signal 107 is divided into two signal components by the first and second signal filters 14, 16 which adjust the signal spectra and provide sufficient amplification necessary for the channels of an A/D converter which is integrated on the same silicon chip as the micro-controller 18. The first and second signal components are fed into a pair of high-pass filter circuits 108, 111. The first of these high-pass filter circuits 108 includes a resistor R12 and a capacitor C5 the values of which are set such that the circuit 108 differentiates the signal. The second high-pass filter circuit 111 also has a resistor Rl 1 and a capacitor C4 but does not differentiate the signal. In the present embodiment, resistors R12 and Rl 1 are set to 100 K ohm and 1 M ohm respectively, and capacitors C5 and C4 set to 47 nF and 220 nF respectively.
From signal theory it is known that only the QRS complex of the ECG signal need be considered to obtain precisely measured R-R intervals. The frequency spectrum of the QRS complex contains frequencies in the range of around 20 to 100 Hz. Undesirable signals such as cardiac artefacts have a 1/f-type spectrum with their main components being located below 20 Hz. It is therefore useful to process the rough ECG signal using a high-pass filter 108 before A/D conversion. In practice a cut-off frequency of around 34 Hz is used.
Filtering and amplifying of the first and second signal components 109, 112 of the ECG signal are performed by OP 295 single-supply micro-power operating amplifiers 110, 113 with "rail- to-rail" capability. These operational amplifiers 110, 113 allow the A/D converter reference voltage to be used as an operating amplifier power supply with no loss in A/D converter input range. Both signal, branches 114, 116 act as an low-pass filters, the cut-off frequencies determined by resistor Rl 0 and capacitor C3, and resistor Rl 5 and capacitor C6, respectively. In the present embodiment, resistors R10 and Rl 5 are set to 100 K ohm and capacitors C3 and C6 are set to 2.2 nF. These low pass filters eliminate noise and radio frequency signals which are greater than the ECG signal spectrum. Resistors R8 and R13 determine the first and second channel gain for the signal branches. Resistors R9 and R14 are used for the differential amplifier 104 input bias current compensation and to avoid unwanted DC level shift in output signal. In the present embodiment, resistors R8, R13, R9 and R14 are respectively set to 3 K ohm, 1 K ohm, 1 M ohm, and 100 K ohm.
The overall operations of the transmitter 10 are controlled by a micro-controller 18. In the described embodiment a PIC16C711 micro-controller is used (manufactured by Microchip). This is a low power device is based on "Dual Bus Harvard RISC" architecture and provides high computational power relative to its power consumption. The micro-controller 18 has four integrated A/D converter channels, only three of which are used, which allow direct connection to the first and second signal filters 14, 16 and thus significantly reduce the complexity of the transmitter circuitry. The main functions performed by the micro-controller 18 are sampling and A/D conversion of the ECG signal from both signal branches.
The sampling rate for the first signal component 109 of the ECG signal is around 1000 Hz. This corresponds to a time interval of around 1ms which is necessary to obtained accurate R-R intervals from the ECG signal. The sampling rate for the second component 112 may be set at somewhere between 100 and 500 Hz. Further functions performed by the micro-controller 18 include coding of measured R-R interval values (first signal component), ECG signal samples (second signal component) and battery status into a serial data stream using a dedicated transfer protocol prior to telemetric transfer, charging control of internal lithium-ion rechargeable battery (two-stage CC and CV charging method), diagnostics and service functions.
As described above, the recorded ECG signal is amplified and digitally processed directly in the transmitter 10 (on the patient's body) and an instantaneous heart rate (R-R intervals) precisely derived. This information together with the raw ECG signal and battery status information are encoded by a transfer protocol into a serial code (suitable for optimal narrowband radio channel exploitation) and transmitted in the UHF band by the radio transmitting module 20 and transmission ariel 22. The transmitted signal is in digital form. Referring now to Figures 2 to 5, the present invention incorporates a technique known as diversity reception which eliminates fading of a transmitted signal. Fading is caused by multipath wave propagation, which results from reflection and absorption of transmitted radio waves between a transmitter and a receiver. Multiple waves received by an ariel will be added together according to the principle of vector summation. This means that in certain conditions (when the waves are in opposite phase) the vector sum may be close to zero and transferred information may be lost.
An effective way of overcoming this problem is to use two spaced apart antennae 32, 34 which increases the probability of a good signal level being received in one antenna if the signal received in the other antenna suffers interference. Summing signals from the two antennae is likely to encounter the problem of out of phase signals cancelling each other out. The signals received by the two receiver antennae 32, 34 are processed by two separate receiving modules 36, 38 and are then switched by the analogue switching circuit which determines the best quality signal. The selected signal is processed in a data recovery circuit (slicer) which separates the digital data from analog signal.
The transmitted signal is received by one or both of a pair of receiver antennae 32, 34 which form part of a remote receiver 30. The receiver antennae 32, 34 operate under diversity reception to ensure that the best quality signal available is received by the receiver 30. he received signal (s) pass through a pair of receiver module 36, 38 which pass the signal(s) to a receiver micro-controller 44 either directly or via an analogue switch 40 and a data recovery unit 42. Data processed by the micro-controller 44 by a suitable connector 46 to and from a PC. In the described embodiment the connector 46 is an RS232-type interface.
DESCRIPTION O PEAK FINDING:
To measure the R-R intervals accurately it is necessary to correctly identify each consecutive R wave peak and measure the time at which each R wave peak occurs. However, R wave peak identification can become complicated as local maxima/minima occur for Q and S wave peaks during each heartbeat as shown in Figure 4. Noise peaks and troughs associated with ECG signals can also lead to further maxima/minima. This is particularly true of very noisy ECG signals which are obtained when the subject is moving or exercising.
As shown in Figures 4 and 5, although there is only one true R wave peak for each heartbeat there will be at least three turning points result from taking the first derivative of the ECG signal. Thus, additional test conditions are required to eliminate false identification.In the present invention, these test conditions are set up according to characteristic Rpeak properties to provide reliable identification of R wave peaks even in a noisy ECG signal.
The method of R wave peak identification will now be described with reference to Figures 4 and 5. The wave peaks and troughs of the PQRST complex are labelled P, Q, R, S, and T. Figures 4 and 5 show respectively the raw ECG signal before and after differentiation. Thus, Q', R1, and S' are the turning points or first moments associated with the Q, R, and S peaks respectively. Figure 5 shows only the QRS portion of the PQRST complex as the P and T waves are suppressed by the first high-pass filter 108, the differentiating circuit . In Figure 5 the time and voltage scales have been magnified for clarity. The y-ordinate which does not represent millivolts but a numeric value resulting from A/D conversion. A nominal zero level of 128 has been adopted.
The method of the present invention is more flexible than existing methods as it takes into consideration the fact the shape of the QRS complex can vary considerably which results in variations in the amplitude of the Rpeak. This can be the case if the electrodes are incorrectly positioned on a subject. This often results in the R and/or S peaks increasing in amplitude. However, the dependence of the pre-processed signal amplitude on the whole QRS complex negates the effect of ECG electrode positioning as the amplitude difference between the global maximum and the global minimum in the differentiated signal is practically constant for any ECG signal shape.
The method and apparatus of the present invention utilise the following points of interest in a differentiated ECG signal as illustrated in Figure 5:
1) the temporal position of the maximum amplitude, associated with the maximal positive ECG slope (Tmax);
2) the temporal position of a turning point associated with an R peak (Tdowα);
3) the maximum amplitude (MAX); and
4) the minimum amplitude, associated with maximal negative ECG slope (MTN).
The occurrence of points 3) and 4) is stored in a temporary memory. These points do not relate to local maximum or minimum but global extremes computed on a time interval defined by the turning points associated with Q, R, and S peaks and troughs. The global maximunVminimum searching algorithm of the present invention always starts in response to the occurrence of a first turning point i.e. Tup. This eliminates false identification of local extreme, because multiple local extremes can appear on the QRS wave as a result of signal noise. The zero of the y-ordinate in Figure 5 relates to a zero signal level or isoelectrical line of the A/D converter i.e. thel28 value corresponds to the zero signal level.
Following from above, each downward zero crossing point could potentially be an R peak occurence. In the present method, these times are written to a memory as a preliminary R peak occurrence or preliminary fiducial point. The present method uses further computational checks to verify each preliminary R peak occurrence as a true R peak.
Before the R wave peak associated with one heartbeat can be positively identified the whole QRS complex for that heartbeat needs to be considered. The R wave peak is confirmed only after the signal minimum 4) or MTN has been identified. The algorithm used in the present invention tests the peak-peak amplitude of each sinusoidal-like waveform (MAX - MTN) to determine whether it exceeds a predetermined limit (AMP). If the amplitude is greater than the predetermined limit the algorithm moves to a further test condition. If the amplitude condition fails, the potential R peak or fiducial point is rejected and the algorithm continues searching for the true R peak
If the amplitude comparison is successful, the next step of the algorithm considers the time interval between 3) the signal maximum or Tmax and 2) the downward zero crossing Tdown. This interval, Tdown-Tmax, is directly related to the shape of the QRS complex. If the calculated time difference is less than a predetermined value Qlim, a false R peak or fiducial point is registered. Such false points are usually caused by myopotential noise. If the time condition test fails, the preliminary fiducial point is rejected and the algorithm continues searching for a true or valid R peak.
If both amplitude and time conditions are satisfied, the potential fiducial point is passed to another part of the memory for storage as a fiducial point. However, one further check is conducted to ensure that an a valid R wave peak has been identified. This additional checking condition is invoked to prevent R-R intervals shorter than a predetermined minimum value (M3NRR) (usually around 100 msec) being recorded. In practice, even a heart beating at 160 times a minute would equate to an R-R interval of approximately 375 ms and therefore intervals of the order of 100ms are physically impossible. Moreover, R-R intervals shorter than 100 msec would result in data processing problems. For these reasons, supposed R wave peaks occurring within such a short time interval after the last positive identification are rejected.
If the next identified peak takes longer than the set time period (MTNRR) to appear it is accepted as a fiducial point. Once the next fiducial point has been accepted as an R wave peak, the R-R interval value is computed as a difference between current R peak or fiducial point time and the previous fiducial point time (stored in memory from last cycle). The computed value is stored and then passed on for transmission purposes.
In the case of another fiducial point occurrence during (MTNRR), the current fiducial point value is discarded, replaced by the next identified fiducial point value and the waiting time counter reset. This final check prevents data congestion during data transmission and identification of false R-R values (artefacts). Such artefacts occur just before real beats and the fiducial point shifting described above can remove these undesirable effects.
As described above, the test conditions consist of comparing the sine-like wave peak-peak. amplitude (MAX - MEN) with a predetermined value. However, in practice the ECG amplitude changes from beat to beat, which means that the amplitude of the R wave peak may vary continuously. This is particularly true if a subject is moving while an ECG is being recorded. Thus, use of a fixed value or threshold for comparing R wave peaks can be unreliable for positive fiducial point identification. A low threshold can result in false positive beat identification (noise will be interpreted as heartbeat) and a high threshold value may result in false negative identification error (heartbeats missing). Methods based on solely on comparing the amplitude against a set threshold have proven to be unreliable, often resulting in identification of false R wave peaks, lost R wave peaks and a lack of precision in the R-R intervals evaluated.
In order to improve the accuracy of R wave peak identification yet further, the present invention uses a dynamic or adaptive threshold which changes periodically. The amplitude of the last successfully identified wave is designated as the threshold. This threshold value is periodically decremented at set time intervals ΔT. The values of the dynamic threshold are given by following geometric progression:
*n-l a„ n-1
Where a„ is the threshold value in present step and an is the threshold value in the previous step. The quotient of this geometric progression is:
q = 1 - 1/8 = 7/8
When plotted against time the values of the progression form a curve which approximates a decaying exponential function: u(t) = U0* exρ(t /τ)
The initial value U0 represents the previous wave amplitude and it can be deduced that the equivalent time constant t is given by following expression:
τ = ΔT/ ln(l/q)
Where τ is time constant of exponentially falling curve, ΔT is time period of the decrementing algorithm (period of iteration) and q is the progression quotient which can be expressed more usefully by the following equation:
q = l - l / D
D is a dividing factor incorporated into the software of the micro-controller 18, and which is set in the source code before compilation. In practice, the micro-controller software divides the last stored threshold value by the dividing factor D and the new threshold value is obtained by subtraction. The dividing factor D value is restricted to 2n numbers only due to limited mathematical capabilities of the micro-controller.
The exponentially falling dynamic threshold prevents false beat identification as the probability of beat occurrence increases in the time, counted from the last heartbeat occurrence moment, corresponding to the falling thresholds. On the contrary, a high level of dynamic threshold just after the heartbeat prevents disturbing noise pulses being falsely identified. From experimental measurements, the dynamic threshold algorithm has been refined such that the threshold decrementation process is delayed for a period, TbIock = 100 msec, to eliminate high level noise peaks just after an identified QRS complex.
All of the R-R interval measuring algorithm parameters (dynamic threshold behaviour, time restricting condition etc.) can be selected before source code compilation to optimise the algorithm performance. FLO DIAGRAM DESCRIPTION:
In the present invention, the micro-controller incorporates a 16 bit free-running timer, having a timer clock fremjencv is 1000 Hz. The timer is incremented once every millisecond, thus the clock rate 1000 Hz and 16 bit capacity of the timer means that it overflows roughly every 64 seconds.
This timescale is much longer than any time interval between features which appear in the ECG signal. Therefore no absolute timescale is required with all relevant time points being identified and stored relative to one another rather than relative to a set start or end point. In particular, the time intervals between points of interest are calculated by relative time point subtractions. There is no problem even when measured interval is falling to timer overflow moment because of binary mathematics properties (MODULO 216computing). Mathematical operations in 16-bit registers automatically perform MODULO operation without any additional programming. The instant timer value is labelled as T in the flow diagram of Figure 6.
The main algorithm loop always begins using a signal which has been differentiated and has then undergone A/D conversion. In each cycle, the algorithm hunts or searches for all of the important points in the differentiated QRS waveform i.e. global maximum (MAX), global minimum (MTN), a downward zero crossing (Tdown), and the temporal position of the global maximum (Tmax). In addition to the main amplitude and time tests conducted on the differentiated QRS waveform the main loop also contains tests which determine if particular actions need. to be performed at particular times i.e. sending R-R time interval values for telemetric transmission or controlling the dynamical threshold decay.
The global maximum searching algorithm permanently compares the current sample value SAMPLE(n) to a value MAX stored in a memory location. If the current sample is greater than the stored MAX value then the stored value is replaced by the current sample (i.e. MAX is updated).The initial value given to the value MAX is reset before each maximum searching cycle.
In each cycle, the nominal zero value is first encountered at Tup (first upward zero crossing). This can be seen in Figure 5. It also useful to note that the global maximum MAX must be located on the curve segment between Tup( first upward zero crossing) and Tdown (the downward zero crossing) due to the nature and shape of the differentiated QRS complex. As the searching cycle progresses the algorithm is incremented and each successive temporary maximum is stored (both its value MAX and time of occurrence Tmax) until the global maximum value is reached. The maximum values are recorded between points 1) and 3) on Figure 5.
The search cycle for the minimum begins after value Tdown is reached and the search for the global minimum is performed in a manner similar to that described above.The memory location MIN is updated every time the current sample SAMPLE(n) is less than a stored value
MEN.
The minimum searching algorithm is reset (nominal zero of 128 is stored to memory location MTN) when Tdo^,n (downward zero crossing point) is encountered. The difference compared to the maximum searching procedure is that the algorithm is not waiting until a new upward zero crossing is reached and a definitive or global mirώnum is found. Instead, the algorithm checks every temporary minimum to see whether that point meets the amplitude condition (MAX - MTN > AMP). When the amplitude condition is met then time condition (Tdown - Tmax > Qlim) is checked and if both are satisfied (i.e. a beat has been detected) then following actions are performed in the micro-controller:
- the new threshold value AMP = MAX - MIN is computed and stored;
- the time of a threshold decay is pre-set to a DECRTIME variable (details follow below);
- the delay time at which R-R interval data will be sent to the transmission procedure is pre-set to a SENDTEME variable (SENDTEVIE = T + MTNRR);
- a sending flag SENDFLAG is set; and
- an LED on the exterior or the transmitter lights to indicate positive detection of an R peak Note that this procedure occurs initially when the temporary detected minimum is sufficiently low (amplitude condition is satisfied) and thereafter each time a lower minimum is detected (during curve falling). As a result of this fact, all values which are saved/pre-set will correspond to true (global) minimum of analysed curve (because every "deeper." minimum causes previously stored values discarding and replacing by new ones). Thus, the new initial dynamic threshold value will be correctly set to a value which is equal to the difference between the global maximum and the true global minimum in spit of the fact that the beat confirmation (amplitude and time condition fulfilment) has been done before reaching the absolute minimum.
A more logical approach would seem to be to simply detect a global minimum i.e to stop when the next upward zero crossing point is reached, and to then apply amplitude and time conditions. However, in practice this solution is not as effective as the negative wave of the analysed curve may persist for a relatively long time below zero level. Thus, if the algorithm waited for the upward zero crossing (S' in Figure 5), an unacceptable beat identification delay would occur and consequently the telemetry transfer would not work properly.
The downward zero crossing is identified when the current sample SAMPLE(n) is zero or drops below value 128 (zero voltage level). Applying only this condition is not sufficient as all samples below the nominal zero level of 128 will be considered to be downward zero crossing. To overcome this an additional condition is used, namely using the last sample SAMPLE(n-l).
If SAMPLE(n-l) lies above the nominal zero level and SAMPLE(n) is equal to or less than the nominal zero level then this indicates a downward zero crossing situation. SAMPLE(n) is then saved as a downward zero crossing time Tdown. In practice, there may be some inaccuracy due to the sampling rate and the resolution of the A/D converter which means that the zero crossing point is not always 128 on the A D converter range. However, fluctuations in the absolute value of the zero crossing point on the A/D converter range have no effect on the final accuracy of the R-R time interval, which always has an accuracy of +/- 0.5 msec (given by sampling rate). Checking the last sample value SAMPLE(n) for zero crossing identification eliminates the need for calculation of a second derivative (which for the R peak would be negative) of the ECG signal.
The downward zero crossing point is used as a preliminary fiducial point and also serves for minimum searching algorithm reset as described above. The downward zero crossing point 5 represents a local maximum and therefore is a potential fiducial point or R peak. The upward zero crossing point is detected in a similar manner (current sample is zero or above zero line and the last sample is below zero line) and it is used for maximum search algorithm reset purpose only.
l o The R-R interval value is finally calculated and sent for transmission if no new fiducial point is detected during a time period MINRR. If such a new fiducial point is detected during the MINRR time period a previously saved fiducial point will be replaced by this new one and new time period MTNRR will start. This methodology is ensured by the following "send time" test:
15 -when current time T reaches the value stored as the SENDTLME variable the following actions are performed:
- R-R value is calculated as a difference between new and old (saved from last beat ientification) fiducial points
- old fiducial point in memory is replaced by a new one (for next R-R calculation) 0 - the SENDFLAG is reset to prevent multiple sending and notifies the sending process
(process lunning on the background) about R-R value transfer request
- the yellow (beat indicating) LED is switched off
The dynamic threshold decay is performed periodically (at intervals of time length DeltaT) in the following manner: 5 - the threshold derogation test checks if the current time T (timer value) has reached the decrement time point, which is saved in memory location DECRTHME.; if the answer is "yes", anew value for the dynamical threshold is calculated according to the formula: AMP = AMP - AMP/8
and a new decrement time point is calculated and saved to the DECRTIME variable:
DECRTIME = T + deltaT
However, there is one exception which is just after the beat detection, a new threshold initial value is calculated, because the DECRTIME variable is pre-set to DECRTIME = T + Tblock
(see the explanation paragraph of minimum identification algorithm). Thus, the dynamical threshold decay is blocked for time period Tblock.
The current sample SAMPLE(n) is saved at the end of whole loop as SAMPLE(n-l) for next cycle usage.

Claims

Claims:
1. A method for identifying an R peak in the QRS complex of an ECG signal, the method comprising the steps of: identifying a maximum and a minimum in the first derivative of the ECG signal; determining an amplititude difference between the maximum and minimum, identifying a first time corresponding to the maximum; identifying a second time corresponding to a point between the maximum and the minimum at which the first derivative of the ECG signal is substantially zero; determining a temporal difference between the first time and the second time; and identifying the point as a potential R peak if the amplitude difference is greater than a predetermined amplitude value and the temporal difference is greater than a predetermined time value.
2. A method as claimed in claim 1 further comprising the steps of: setting a time period after the identification of the potential R peak; and rejecting the potential R peak if any further potential R peak occurs within the period.
3. A method as claimed in claim 1 further comprising the steps of: setting a time period after the identification of the potential R peak; and, accepting the potential R peak as a valid R peak if no further potential R peak occurs within the period.
4. A method as claimed in claim 3 further comprising the step of calculating the time interval between the valid R peak and the last recorded R peak to evaluate an R-R interval.
5. A method as claimed in any of claims 2 to 4 further comprising the step of setting a time period of approximately 100 ms.
6. A method as claimed in any preceding claim further comprising the step of setting the current calculated amplitude difference to be the next predetermined amplitude value.
7. A method as claimed in any preceding claim further comprising the step of decrementing the predetermined amplitude value based on a previous value.
8. A method as claimed in any preceding claim wherein the step of evaluating the first derivative of the signal is carried out using hardware
9. A method as claimed in any preceding claim wherein the step of identifying fiducial points is carried out using a microprocessor.
10. Data processing apparatus for identifying an R peak in the QRS complex of an ECG signal, the apparatus comprising: a microprocessor; a memory in communication with the microprocessor and storing instructions executable by the microprocessor to: identify a maximum and a minimum in the amplitude of the first derivative of the ECG signal; determine an amplitude difference between the maximum and minimum; identify a first time corresponding to the maximum; identify a second time corresponding to a point between the maximum and the minimum at which the first derivative of the ECG signal is substantially zero; determine a temporal difference between the first time and the second time; and identify the point as a potential R peak if the amplitude difference is greater than a predetermined amplitude value and the temporal difference is greater than a predetermined time value.
11. Data processing apparatus as claimed in claim 10 in which the microprocessor: sets a time period after the identification of the potential Rpeak; and rejects the potential R peak if any farther potential R peak occurs within the period.
12. Data processing apparatus as claimed in claim 10 in which the microprocessor: sets a time period after the identification of the potential R peak; and, accepts the potential R peak as a valid R peak if no further potential R peak occurs within the period.
13. Data processing apparatus as claimed in claim 12 in which the microprocessor calculates the time interval between the valid R peak and the last recorded R peak to evaluate an R-R interval.
14. Data processing apparatus as claimed in any of claims 11 to 13 in which the microprocessor sets a time period of approximately 100 ms.
15. Data processing apparatus as claimed in any of claims 10 to 14 in which the microprocessor sets the current calculated amplitude difference to be the next predetermined amplitude value.
16. Data processing apparatus as claimed in any of claims 10 to 15 in which the microprocessor decrements the predetermined amplitude value based on a previous value.
17. Data processing apparatus as claimed in any of claims 10 to 16 in which the first derivative of the ECG signal is evaluated out using hardware
18. A method substantially as described hereinbefore with reference to the accompanying drawings.
19. Data processing apparatus substantially as described hereinbefore with reference to the accompanying drawings.
PCT/GB2004/001187 2003-03-28 2004-03-18 Method and apparatus for for identifying features in an ecg signal WO2004084722A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
GBGB0307212.1A GB0307212D0 (en) 2003-03-28 2003-03-28 Method and apparatus for identifying features in an ECG signal
GB0307212.1 2003-03-28

Publications (1)

Publication Number Publication Date
WO2004084722A1 true WO2004084722A1 (en) 2004-10-07

Family

ID=9955759

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/GB2004/001187 WO2004084722A1 (en) 2003-03-28 2004-03-18 Method and apparatus for for identifying features in an ecg signal

Country Status (2)

Country Link
GB (1) GB0307212D0 (en)
WO (1) WO2004084722A1 (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006113698A1 (en) * 2005-04-18 2006-10-26 Medtronic, Inc. Method and apparatus for identifying oversensing using far-field intracardiac electrograms and marker channels
WO2007073464A1 (en) * 2005-12-15 2007-06-28 Ivy Biomedical Systems, Inc. A system and method for triggering a device based on an electrocardiogram signal
WO2007130219A1 (en) * 2006-04-27 2007-11-15 Medtronic, Inc. Peak data retention of signal data in an implantable medical device
CN102379694A (en) * 2011-10-12 2012-03-21 中国科学院苏州纳米技术与纳米仿生研究所 Electrocardiogram R wave detection method
EP2753236A4 (en) * 2011-09-08 2015-07-01 Apn Health Llc R-wave detection method
CN105286857A (en) * 2015-09-29 2016-02-03 北京航空航天大学 R wave rapid detection method adaptive to electrocardiogram waveform pathological change
CN107997759A (en) * 2017-10-27 2018-05-08 北京康博众联电子科技有限公司 The detection method of QRS complex, storage medium and computer equipment in electrocardiosignal
RU2791006C1 (en) * 2019-05-24 2023-03-01 Автономная некоммерческая образовательная организация высшего образования "Сколковский институт науки и технологий" System and method for automated electrocardiogram analysis and interpretation
US11944819B2 (en) 2020-07-09 2024-04-02 Medtronic, Inc. Analyzing ECAP signals

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114504326B (en) * 2022-01-17 2023-07-18 电子科技大学 Binary amplitude coding method for electrocardiosignal

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3939824A (en) * 1973-10-09 1976-02-24 General Electric Company Physiological waveform detector
EP0003567A1 (en) * 1978-02-09 1979-08-22 Hellige GmbH Interference suppression of QRS detector by automatic threshold leveling
CZ278480B6 (en) * 1992-08-31 1994-01-19 Fakulta Telesne Kultury Univer Comparator network of an apparatus for detection and transmission of a heart pulse period
US6070097A (en) * 1998-12-30 2000-05-30 General Electric Company Method for generating a gating signal for cardiac MRI
WO2002100264A2 (en) * 2001-06-12 2002-12-19 Wolf Tamas Method and apparatus for electrocardiologic self-examination

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3939824A (en) * 1973-10-09 1976-02-24 General Electric Company Physiological waveform detector
EP0003567A1 (en) * 1978-02-09 1979-08-22 Hellige GmbH Interference suppression of QRS detector by automatic threshold leveling
CZ278480B6 (en) * 1992-08-31 1994-01-19 Fakulta Telesne Kultury Univer Comparator network of an apparatus for detection and transmission of a heart pulse period
US6070097A (en) * 1998-12-30 2000-05-30 General Electric Company Method for generating a gating signal for cardiac MRI
WO2002100264A2 (en) * 2001-06-12 2002-12-19 Wolf Tamas Method and apparatus for electrocardiologic self-examination

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7567835B2 (en) 2005-04-18 2009-07-28 Medtronic, Inc. Method and apparatus for identifying oversensing using far-field intracardiac electrograms and marker channels
US8792971B2 (en) 2005-04-18 2014-07-29 Medtronic, Inc. Method and apparatus for identifying oversensing using far-field intracardiac electrograms and marker channels
US8386024B2 (en) 2005-04-18 2013-02-26 Medtronic, Inc. Method and apparatus for identifying oversensing using far-field intracardiac electrograms and marker channels
WO2006113698A1 (en) * 2005-04-18 2006-10-26 Medtronic, Inc. Method and apparatus for identifying oversensing using far-field intracardiac electrograms and marker channels
US7920912B2 (en) 2005-12-15 2011-04-05 Ivy Biomedical Systems, Inc. System and method for triggering a device based on an electrocardiogram signal
WO2007073464A1 (en) * 2005-12-15 2007-06-28 Ivy Biomedical Systems, Inc. A system and method for triggering a device based on an electrocardiogram signal
US7765088B2 (en) 2006-04-27 2010-07-27 Medtronic, Inc. Peak data retention of signal data in an implantable medical device
US7359837B2 (en) 2006-04-27 2008-04-15 Medtronic, Inc. Peak data retention of signal data in an implantable medical device
WO2007130219A1 (en) * 2006-04-27 2007-11-15 Medtronic, Inc. Peak data retention of signal data in an implantable medical device
EP2753236A4 (en) * 2011-09-08 2015-07-01 Apn Health Llc R-wave detection method
CN102379694A (en) * 2011-10-12 2012-03-21 中国科学院苏州纳米技术与纳米仿生研究所 Electrocardiogram R wave detection method
CN105286857A (en) * 2015-09-29 2016-02-03 北京航空航天大学 R wave rapid detection method adaptive to electrocardiogram waveform pathological change
CN107997759A (en) * 2017-10-27 2018-05-08 北京康博众联电子科技有限公司 The detection method of QRS complex, storage medium and computer equipment in electrocardiosignal
RU2791006C1 (en) * 2019-05-24 2023-03-01 Автономная некоммерческая образовательная организация высшего образования "Сколковский институт науки и технологий" System and method for automated electrocardiogram analysis and interpretation
US11944819B2 (en) 2020-07-09 2024-04-02 Medtronic, Inc. Analyzing ECAP signals

Also Published As

Publication number Publication date
GB0307212D0 (en) 2003-04-30

Similar Documents

Publication Publication Date Title
FI114282B (en) Method, Arrangement and Heart Rate Monitor for Heartbeat Detection
CN107041743B (en) Real-time R wave detection method for electrocardiosignals
EP1626657B1 (en) Patient monitor
US7379770B2 (en) Devices and methods for heart rate measurement and wrist-watch incorporating same
EP2628444B1 (en) Pulse period computation device and bio-sensor provided with same
AU638648B2 (en) Wrist worn heart rate monitor
CN106028918B (en) Body-worn system for continuous non-invasive measurement of vital signs
US5738104A (en) EKG based heart rate monitor
US6773397B2 (en) System for processing signal data representing physiological parameters
US8167800B2 (en) Apparatus and method for processing pulse waves
US20180103855A1 (en) Method and Apparatus for Heart Rate Monitoring
EP2589332A1 (en) System and method for the analysis of electrocardiogram signals
CN111513706B (en) Method and device for detecting electrocardiosignals containing abnormal R waves
US20140128754A1 (en) Multimodal physiological sensing for wearable devices or mobile devices
KR101827522B1 (en) Apparatus and method for measuring heart rate through non-contact
CN111481190A (en) Robust heart rate measurement method and device based on two-path PPG signal
WO2004084722A1 (en) Method and apparatus for for identifying features in an ecg signal
KR101514151B1 (en) Noise robust algorithm for ppg signal measurement
CN116849669A (en) Electrocardiosignal signal processing system and defibrillator
JP2019013479A (en) Heart beat and respiration measuring system, and heart beat and respiration measuring method
EP1641388B1 (en) Devices and methods for heart-rate measurement and wrist-watch incorporating same
EP1774907A1 (en) Method and device for measuring heart rate
KR101357098B1 (en) Method for detecting and correcting of bio-signal of walking surport device user
EP1608265A1 (en) Apparatus for processing an ecg signal
US20200093391A1 (en) Pulse discrimination device and electrocardiogram analyzer

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): BW GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
122 Ep: pct application non-entry in european phase