Classifications

• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
  • A61B5/024—Detecting, measuring or recording pulse rate or heart rate
  • A61B5/02405—Determining heart rate variability
• • 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
  • A61B5/24—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
  • A61B5/364—Detecting abnormal ECG interval, e.g. extrasystoles, ectopic heartbeats

Definitions

• the present invention relates to the field of digital processing of an analog bio-electric signal, which is characteristic of the heart rate of a living being, and which is designated in the present text by the terms "heart signal”; it is preferably, but not exclusively, an electrocardiographic signal (ECG).
• ECG electrocardiographic signal
• the main objects of the invention are a method and a device for filtering an RR series, obtained by sampling a cardiac signal.
• the heart of a living being contracts automatically very regularly like a metronome, under the action of the sinus node which generates an independent nerve impulse, and thereby - even causes a spontaneous contraction of the heart muscle.
• the heart is not isolated, however, but is linked to the Autonomous Nervous System (ANS), through the parasympathetic and sympathetic systems.
• ANS Autonomous Nervous System
• This autonomic nervous system influences the activity of the heart: the sympathetic system accelerates the heart rate, while the parasympathetic system slows it down.
• the heart undergoes influences from the autonomic nervous system, which in particular allows the organism of a living being to adapt the heart rate according to its needs, within however reasonable limits.
• a known invasive technique consists for example in using a blood pressure sensor connected to a catheter introduced into an artery.
• non-invasive methods there is for example the use of an infrared pulse sensor, or the acquisition of an electrocardiographic signal (ECG) by means of an electrocardiograph.
• ECG electrocardiographic signal
• This latter method of acquiring an ECG signal is in practice the most commonly used to date, because in addition to its non-invasive nature, it advantageously makes it possible to obtain a more precise signal than that obtained for example by means of a sensor. infrared pulse.
• the ECG signal is in known manner consisting of a succession of electrical depolarizations, the appearance of which is shown in FIG. 3 appended.
• the P wave which corresponds to the depolarization of ear cups, has a low amplitude, and a dome shape.
• the PQ space translates the atrioventricular conduction time.
• the QRS complex reflects ventricular contraction, and the T-wave reflects ventricular repolarization.
• the peak R is considered as a marker of the ventricular systole, that is to say of the "heartbeat".
• the R wave being most often the thinnest and widest part of the QRS, it is generally used to punctually locate the heartbeat with very good precision, in practice of the order of a thousandth of a second.
• the time interval between two successive R waves characterizes precisely the time separating two successive heart beats; this is the period of the ECG signal, and the reverse of this period gives the instant heart rate.
• RR series representing the evolution over time of the instantaneous heart rate
• the ECG signal which is an analog signal (analog / digital conversion of the ECG signal) is sampled, and the sampled digital ECG signal is processed, automatically detecting the R waves in this digital signal.
• An RR series is thus usually made up of a plurality of samples
• each sample (RR,) corresponding to the time interval, separating two successive R waves from the ECG signal.
• RR series is not limited not to the aforementioned particular definition based on the R waves of an ECG signal, but is defined more generally in the context of the present invention as a series of several so-called samples (RR,), obtained after sampling of an analog heart signal which is characteristic of the heart rate, each sample (RR,) characterizing the time interval between two successive heart beats.
• the RR series resulting from a cardiac signal, and for example from an ECG signal, with a view to its spectral analysis, is usually transposed in the frequency domain, using different known methods.
• the most commonly used method is to compute the discrete Fourier transform of the RR series.
• Another known method is to calculate the Wigner-Ville quadratic transform of the RR series.
• disturbances in the cardiac signal and in particular in an ECG signal, induce, in the RR series resulting from this cardiac signal, abrupt variations of short duration commonly called artefacts.
• the disturbances, which cause artifacts in the RR series can be physiological and intrinsically linked to a momentary dysfunction of the cardiac system; it is for example an extrasystole.
• These disturbances can also be external and not linked to the functioning of the cardiac system; it is for example a movement of the patient briefly altering the measurement signal.
• Artifacts in an RR series can result in a single erroneous sample or in a plurality of successive erroneous samples.
• an artifact in the RR series can be assimilated to a Dirac impulse, and results in- the field frequency by a wideband rectangular continuous spectrum. Consequently, on the assumption that one would transpose into the frequency domain (by Fourier transform or other) an RR series, without first taking special precautions, the presence of artefacts in the RR series would result in the frequency domain by obtaining a frequency spectrum of the highly disturbed RR series, of broadband rectangular shape, masking the spectrum of the real signal.
• the procedure is as follows.
• the analog cardiac signal for example an ECG signal
• an RR series of this signal is automatically constructed, and stored in memory, so as to make an analysis thereof.
• spectral delayed compared to the recording of the cardiac signal. This delayed spectral analysis is carried out as follows. In a first step, an operator performs manual filtering of the RR series.
• the RR series saved in memory is displayed on a screen for the operator, who visually detects each erroneous sample characteristic of an artefact; the operator manually selects one or more "clean" portions of the RR series free of artifacts.
• the frequency transposition for example by Fourier transform, is calculated only on the “clean” portion (s) selected manually by the operator.
• a major drawback of the aforementioned filtering method is that it requires human intervention to detect the artefacts, and especially to select the "clean" portions of the RR series, this which makes it tedious and relatively long to implement; this method is therefore unsuitable for processing cardiac signals, over long acquisition durations, and for example acquisition durations of one day or more.
• Another disadvantage of the aforementioned filtering method is that it translates in practice by a suppression of large time portions in the RR series, and therefore does not allow a spectral analysis, over the entire duration of acquisition of this signal. heart.
• the present invention relates to a new filtering process for an RR series which is automatic and makes it possible to overcome all or part of the aforementioned drawbacks.
• the method for filtering a series RR is characterized in that the erroneous sample (s) is detected and automatically filtered in the series (RR), and in that for detecting whether a sample (RR, ) is erroneous, the value of this sample (RR,) of the series is compared with at least one self-adapting threshold which is calculated from (N) samples of the series (RR) taken in a sliding window.
• the filtering of the erroneous samples is obtained by purely and simply deleting these samples from the RR series.
• each erroneous sample or each succession of erroneous samples is filtered by being replaced in the RR series, by one or more corrected samples RR k calculated by linear interpolation . More particularly, it is tested whether the previous sample RR. , is correct, and if not, we attempt a linear interpolation with the RR sample, and we automatically detect that the RR sample is correct or incorrect depending on the result of the interpolation.
• the filtering method of the invention can be implemented in delayed time with respect to the acquisition of the cardiac signal and to the construction of the RR series from the recorded cardiac signal.
• the filtering process is applied to an RR series, the points of which are stored in memory.
• the filtering method of the invention can also be implemented in real time, as the cardiac signal is acquired.
• Another subject of the invention is therefore a method of acquiring and processing an analog heart signal, characteristic of the heart rate, and in particular of an ECG signal.
• This process is known in that the heart signal is recorded, this signal is digitized and an RR series is constructed.
• filtering of the RR series is carried out in real time, as this series is constructed, by implementing the aforementioned filtering method of the invention.
• the invention also relates to a system for real-time acquisition and processing of an ECG signal, which system includes measurement electrodes allowing the acquisition of an ECG signal, and means for processing the ECG signal which include an analog / digital converter for sampling the ECG signal, and a programmed processing unit receiving as input the signal from the converter; the processing unit is programmed to automatically construct, from the signal delivered by the converter, a series RR consisting of a plurality of samples (RR,) respectively defining the time intervals which separate two successive heart beats, and for filter the RR series in accordance with the aforementioned method of the invention.
• a series RR consisting of a plurality of samples (RR,) respectively defining the time intervals which separate two successive heart beats, and for filter the RR series in accordance with the aforementioned method of the invention.
• FIG. 1 schematically represents the main elements of a system for the acquisition and frequency processing of an ECG signal according to the invention
• FIG. 2 is a block diagram of the three main functional modules of the processing software executed by the processing unit of the acquisition system of Figure 1, - Figure 3 shows the wave (PQRST) characteristic of a signal
• FIG. 4 shows an example of digital ECG signal ; obtained after sampling an analog ECG signal
• FIG. 5 represents the RR series constructed from the signal of FIG. 4,
• FIG. 6 to 9 respectively illustrate four types of disturbances which are likely to be present in an RR series, and which are detected and filtered by the filtering algorithm of the invention
• FIG. 10 is a flowchart illustrating the main steps of the main routine of an example of a filtering algorithm according to the invention
• FIG. 1 1 is a flowchart illustrating the main steps of a simplified variant of an interpolation subroutine which is called by the main routine of the figure 10,
• FIG. 12 a flowchart illustrating the main steps of an improved variant of an interpolation subroutine which is called by the main routine of Figure 10.
• Figure 1 an acquisition system and frequent heart rate treatment. This system includes:
• ECG electrocardiographic monitor
• the ECG signal processing means 3 comprise an analog / digital converter 4, and a programmed processing unit 5.
• the input of the converter 4 is connected to the output of the ECG monitor 2, and the output of the converter 4 is connected to a input port of the processing unit 5.
• the processing unit 5 is constituted by a microcomputer, the converter 4 being connected to an RS232 serial port of this microcomputer.
• the electrodes 1 are applied to the body of a patient, and the ECG monitor delivers as usual an analog electrical signal, called the ECG signal, which for each heartbeat, in the form of the signal shown in the figure. 3.
• This ECG signal is digitized by the converter 4, with a predetermined sampling frequency (f), for example equal to 256 Hz.
• the P wave which corresponds to the depolarization of the atria has a low amplitude and a dome shape; the PQ space which translates the atrioventricular conduction time; the QRS complex reflects ventricular contraction, and the T-wave reflects ventricular repolarization.
• the R wave is a marker of the ventricular systole, or of the "heartbeat”.
• the RR interval / figure 4 The RR interval corresponds to the time separating two heartbeats, it is the instantaneous period of the signal, and its inverse gives the instantaneous heart rate.
• the R wave being most often the thinnest and widest part of the QRS, allows to punctually locate the heartbeat with very good precision (of the order of a thousandth of a second).
• the recording of the succession of R waves, from the ECG signal, makes it possible to construct the RR series and to analyze it in the frequency domain.
• FIG. 2 shows the three main functional modules 7, 8, 9 of the digitized ECG signal processing software.
• the first module 7 is supplied with input and in real time with the successive digital data constituting the digitized ECG signal
• the first module 7 is designed to automatically detect each peak R, successive in the digital signal 6, and to construct automatically an RR series from this signal. At output, this module 7 delivers successively over time, the RR points, successive of the RR series.
• each point RR is equal to the time interval ( ⁇ t,) (expressed as a multiple of the sampling frequency f) separating a peak R, from the following peak R, +1 (in another variant it could be the previous peak R, . ,).
• the second module 8 designated “RR filter” is designed so as to automatically implement the filtering method of the invention.
• This module 8 which constitutes the essential part of the system with regard to the invention will be described in detail later. At output, this module 8 delivers in real time, a corrected RR series.
• the third module 9 performs the spectral analysis of the corrected RR series.
• This module 9 being moreover already known per se, it will not be described in detail.
• This module 9 generally calculates the power spectral density of the RR series.
• this module transposes the RR (temporal) series into the frequency domain, for example by calculating the discrete fast Fourier transform of this series, in predefined time windows, weighted by means of a predefined weighting window.
• a predefined weighting window may be a rectangular weighting window, or even for example a Kaiser, Hamming or Bartiett weighting window.
• the calculation time windows can be predefined and fixed, or it can be a calculation time window, of predetermined size, which is dragged over time.
• the spectral analysis is performed on a sliding time window of 256 seconds, applied to the RR series, and subjected to a Kaiser weighting to limit side effects due to windowing.
• this weighted window we calculate the Fast Fourier Transform (TFR) of the RR series, and we obtain the power spectral density curve between 0 and 2hz (in accordance with Shannon's theorem).
• the module 9 also calculates, from the spectral density curve obtained, the spectral powers (areas under the spectral density curve) between limits of predetermined frequencies (possibly adjustable by a user). These spectral power calculations constitute a means of investigation and analysis of cardiac regulation by the Autonomous Nervous System (ANS).
• the module 9 was designed to calculate a low frequency spectral power (PS-BF) over a frequency range between 0.039 HZ and 0.148 Hz, and a high frequency spectral power (PS-HF ) over a frequency range of 0.148 Hz to 0.4 Hz.
• Low frequency spectral power (PS-BF) was used to estimate sympathetic and parasympathetic tone
• high frequency spectral power (PS-HF) was used to estimate parasympathetic tone.
• the analog ECG signal is subjected to different types of disturbances (extrasystoles, patient movements, ...) which induce sudden variations in the RR series giving an incorrect evaluation of the spectral analysis.
• the filtering algorithm of module 8 which is detailed below, makes it possible to remedy this problem, by automatically detecting any aberrant variation in the instantaneous heart rate and reconstructing the RR series by linear interpolation while preserving the real time of recording.
• This type of disturbance is equivalent to a succession of disturbances of the 1st type (error by missing a peak R).
• the filtering algorithm of module 8 is broken down into two routines:
• FIG. 10 represents the general flowchart of this main routine, designated “RR FILTER”,
• FIG. 1 shows the general flowchart of this secondary routine, designated "Interpolation attempt".
• this main routine which allows the automatic detection of erroneous samples of the RR series is the most important of the filtering module 8.
• an RR series can for example have low variability, in which case an erroneous RR point will be all the more easily detected the higher its amplitude; on the contrary, a RR series can have a high variability, in which case an erroneous RR point will be more difficult to detect.
• the filtering algorithm is of the self-adapting type, that is to say calculates detection thresholds whose value depends on the variability of the RR series: the detection limits calculated will be high for RR series with high variability; the calculated detection thresholds will be lower for RR series with low variability. Also, as will appear more clearly below, each detection threshold is not calculated over the entire series, but in a sliding time window comprising a number of predetermined samples (N). Detailed description: The variables in the flowchart in Figure 10 are as follows:
• N configurable integer defining the number of samples used for the calculation of the self-adapting thresholds S1 and S2; The lower the value of N, the finer the detection of erroneous RRs, but the longer the calculation time. It is up to the person skilled in the art to find the best compromise between the computation time and the quality of the detection. In a specific embodiment, given for information only, N was equal to 20.
• RR RR time interval expressed in ms. RR,: RR time interval during processing. RR. ,: previous RR time interval.
• RR C last RR correct time interval.
• M average of the N samples of the sliding window RR.
• ⁇ Standard deviation of the N samples RR of the sliding window.
• the first initialization step (FIG. 10 / step 101) consists in loading into a waiting list, for example of the FIFO type, N first samples RR; A calculation window is thus initialized, having a size of N successive RR samples. In this window of
• each new RR point is treated iteratively (steps 102 to 11) so as to automatically determine the state of this RR point, this state being able to be “correct”, “erroneous”, or “undetermined”.
• the method of determining the state of RR depends on the state of the RR. , previous (test 103 or 109). Case where RR ⁇ is correct (test 103 positive)
• the current sample RR is compared, with two self-adaptive detection thresholds S1 and S2 (test 104), which are calculated from the means (M) and standard deviation ( ⁇ ).
• This self-adaptive thresholding allows a first detection of suspect points, whatever the variability of the RR series.
• the self-adaptive thresholds S1 and S2 are given by the following formula:
• RR C designates the last RR in time which has been characterized as being correct.
• the filtering algorithm tests whether the RRi is wrong or not (step 109).
• the filtering algorithm automatically executes the "interpolation attempt" subroutine. If not, the filtering algorithm checks ( Figure 10 / step 1 1 1) if the current sample RR, is in the interval [S1, S2]. If so, the sample RR is characterized as being "correct”, as well as all the previous samples R,.rod... which were temporarily characterized as being“ indeterminate ”( Figure 10 / step 112). Then, similarly to step 105 previously described, the algorithm slides the window for calculating a sample (entry into the waiting list for the correct sample RR ,, and exit from the most old, i.e. sample R, .N ). The new mean (M) and standard deviation ( ⁇ ) values in this new window are then calculated, with a view to calculating the self-adaptive thresholds S1 and S2 during the following iteration.
• Linear interpolation algorithm (step 121): This algorithm is broken down into two steps. In a first step, this algorithm calculates the corrected points RR k . To this end, it calculates the equation of the line passing through the two points (RR, and RR C ), that is:
• RR d first correct point after the disturbance
• T disturbance time (sum of points to be replaced + RR d )
• RRi b / (1 -a)
• the algorithm performs a correction of the points RR k .
• the RR intervals are expressed in ms. RRs are therefore rounded whole values. These roundings are likely to change the total time of the reconstituted area in proportion to the number of erroneous points.
• the correction consists in calculating the induced difference (sum of the erroneous points - sum of the reconstituted points) and in distributing on each new point a portion of this difference.
• This second step constitutes an improvement, and may not be implementation in a simpler variant.
• This test checks whether the first corrected interpolation sample is equal to the RR C sample with a tolerance of +/- 10% and whether the last corrected interpolation sample is equal to the RR sample ,, with a tolerance of +/- 10%. The test is positive (successful interpolation) when the two above conditions are met.
• FIG. 12 represents a flowchart of an improved variant of implementation of the secondary routine (“attempt at interpolation”).
• this improved variant we do not immediately carry out a linear interpolation between the RR sample and the RR C sample, but we start by performing a linear interpolation between the RR sample and the last point characterized as “indeterminate "(Step 128). If the interpolation is not successful ("testV” negative), we start again with the penultimate sample characterized as "indeterminate” (step 129), etc. It is only in the case where all the attempts at interpolation with the indeterminate samples have failed that an attempt at interpolation is made with RRc (FIG. 12 / step 130).
• RRi is characterized as being "wrong". This technique has the advantage of modifying only a minimum of samples. Indeed, in the case of a successful interpolation, only the erroneous and indeterminate samples occurring after the sample used for this interpolation and before RRi are replaced by the RRk. All other pending indeterminate samples are returned as "correct”.
• an initial test (step 125) is carried out on the number of erroneous points in the queue (erroneous points awaiting reconstruction).
• the variable Nbmax characterizes this maximum number of erroneous points in the queue.
• Her value is user adjustable. It allows to take into account strong accelerations or decelerations of the heart rate (example in the newborn the variations are stronger). Generally, this parameter is used to set the maximum authorized length of a disturbance zone. Beyond this length, all pending and undefined erroneous points are returned as correct.

Landscapes

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

Applications Claiming Priority (3)

Publications (1)

Family

ID=8860576

Family Applications (1)

Country Status (4)

Families Citing this family (9)

Citations (1)

Family Cites Families (5)

• 2001
  • 2001-02-28 FR FR0102760A patent/FR2821460B1/fr not_active Expired - Lifetime
• 2002
  • 2002-02-11 AU AU2002237374A patent/AU2002237374A1/en not_active Abandoned
  • 2002-02-11 EP EP02703688A patent/EP1366428A2/de not_active Ceased
  • 2002-02-11 WO PCT/FR2002/000513 patent/WO2002069178A2/fr not_active Application Discontinuation

Patent Citations (1)

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events