DE102006057882A1 - Electrocardiogram signals evaluation method for atrial fibrillation detection, involves classifying occurred RR-intervals according to preset number of different time intervals, and analyzing time up to repetition of identical intervals - Google Patents

Abstract

The method involves detecting heart beat as a sequence of R-points and recording the heart beat as RR-intervals. The occurred RR-intervals are classified according to a preset number of different time intervals. The time upto the repetition of identical intervals is stored and analyzed. Different repetition times upto new occurrence of long RR-intervals are pre-determined. A counter is assigned to each repetition time such that the number of occurence i.e. finding of identical interval according to repetition time, is stored in the counters. An independent claim is also included for a device for implementing a method for evaluation of electrocardiogram signals.

Description

The The present invention is in the field of medical instrumentation. In particular, the invention relates to methods and devices for the detection of irregular Behavior in the heart rhythm, especially atrial fibrillation.

The Probability for the occurrence of atrial fibrillation (atrial fibrillation, AF) increases with age and is currently in the over 80s age group stated as 6%. A strong increase of this age group until the Year 2050 is to be expected. In addition, the occurrence of AF in epidemiology underestimated there often no symptoms are perceived, which causes about half of the Sufferers are unaware of their illness. The illness brings a significant increase the morbidity and mortality (e.g., strokes) with itself, and occurs in particular frequently in such pre-existing diseases, which indicate a therapy by electrically active implants. The possible early Asked diagnosis improves the chances of recovery as an immediate treatment the prospects of a permanent conversion to sinus rhythm favors. Complicating the correct automatic diagnosis of atrial fibrillation by the differentiation of other arrhythmias, which however one have other clinical relevance of atrial fibrillation or other treatment strategies require.

A Observation of the coordination of atrial and ventricle action as well Atrial IEGM can provide good detection of dual chamber devices Enable atrial fibrillation. However, for example, 50% of single-chamber devices are currently implanted in the ICD area, where - under other reasons of cost - in this regard one rising trend is observed. On the other hand, telemedical Concepts like the "home monitoring" possibilities open, to provide a quick diagnosis of atrial fibrillation, provided that find suitable parameters that reliably recognize a pathological condition can, and that can be easily calculated and transmitted.

A general and applicable in all implants solution to the problem The AF detection in terms of diagnostic assistance exists so far Not. At present, electrically active implants, such as cardiac pacemakers, are used or ICDs, an AF detection for two other tasks. On the one hand, the goal of dual-chamber devices is in the case of atrial arrhythmias as u. a. Atrial fibrillation in a ventricular controlled mode (e.g., DDI) On the other hand, in defibrillators an inadequate shock delivery as a result of quickly being transferred supraventricular Tachycardias are avoided. Both existing procedures are only for a general diagnosis suitable. In the frequently used Einkammergeräten In addition, the atrial signal is generally not available. at Defibrillators without atrial signal is again the discrimination supraventricular Tachycardia according to stability criteria only if and only if the tachycardia criteria exceeded due meets the frequency criteria are, d. H. a routine AF monitoring the nature of things is in both cases not possible.

By a detection of atrial fibrillation in implants, especially Single-chamber devices, could be Significantly shorten the time that passes to the diagnosis, since not the next one anymore clinical examination must be awaited, which is of even greater importance is frequent, as AF without symptoms, and the patient therefore does not perceive his incipient disease. There should be a transfer greater Data volumes from the implant bypassed, and a high computational effort in the implant for This purpose can be avoided by using the application with existing ones Technology without loss the rest functionality to make it realizable.

The successful detection of atrial fibrillation should in particular the Serve purpose in telemedical applications at specific time periods (eg once a day) to transfer appropriate markers, and thus the initiation of a rapid therapy z. B. at new to allow for permanent atrial fibrillation, as newly manifested The sooner this happens, the better the treatment for atrial fibrillation.

The Invention is based on the described prior art. you the task is based on a method and associated system to develop the atrial fibrillation detection that the aforementioned Applications allows and for suitable for use in implants.

These Task is in a method according to the preamble of the claim 1 fulfilled by the characterizing features of claim 1. others Benefits and special embodiments of the method according to the invention are the subject of Dependent claims. The device according to the invention to carry out of the method in portable or implanted devices The subject matter of claim 6. The invention will be described below preferred embodiments with reference to the drawings and the symbols used therein and reference numerals closer explained.

It shows:

1 Definition of RR interval at ECG recordings

2 Detail of a time series of RR intervals

3 Lorenz plot for RR intervals in sinus rhythm with significant contamination of ventricular extrasystoles (VES)

4 Lorenz plot for RR intervals with chronic AF recognizable in the associated ECG

5 Theoretical probabilities for recurrence of an equal length RR interval after 1 to a maximum of 25 heart beats at different values for k (number of occurring different RR intervals)

6 Probabilities for recurrence of an equally long RR interval after 1 to a maximum of 25 heart beats, calculated from different measurement data sets

7 Scheme of a calculation procedure according to the method of the invention (Nassi-Shneiderman diagram)

The distance between two consecutive heart beats is generally defined by the so-called RR interval. This results as a time interval between two significant rashes (so-called R peaks) in the surface electrocardiogram (ECG) or in the intracardiac electrogram, which is measured by means of electrodes through an implant. These characteristic signals represent the contraction of the heart muscle during the course of the beating cycle. 1 shows a section of a typical waveform in a surface ECG with a marked RR interval.

Continuous measurement of R peaks over an extended period provides a sequence of RR intervals, referred to as time series. An excerpt from such a time series is in 2 shown. It shows no quiescent equidistant sequence of R-peaks, ie the interval lengths vary around an average. These fluctuations carry information about the regulatory mechanisms of the cardiovascular system and it is generally believed that they can be used, for example, to detect differences between physiological behavior and pathological changes.

Simple plots of the RR interval above the previous interval for all measured cycles are usually referred to as Lorenz plots (alternatively: Poincaré plots). These provide a two-dimensional graphical representation of the presence of sinus rhythm or interference with the normal sinus rhythm due to the characteristic shapes that occur, such. Of ventricular extrasystoles (shown in 3 ). Atrial fibrillation is also reflected in the Lorenz plot. The 4 shows a Lorenz plot for measured RR intervals in chronic AF.

• – einfache Parameter zur Quantifizierung (z. B. Streuungsmaße senkrecht zur Winkelhalbierenden) sind nicht sehr spezifisch auf AF
• – Mustererkennungsverfahren, die höhere Aussagequalität gewährleisten können, erfordern die Übertragung größerer Datenmengen und dadurch bedingten höheren technischen Aufwand hinsichtlich Speicherung und Übertragung
• – die Übertragung/Auswertung größerer Datenmengen würde in regelmäßigen Abständen erfolgen müssen (frühzeitige Diagnose)

However, the Lorenz plot has a number of disadvantages for practical application:

• - simple parameters for quantification (eg scattering dimensions perpendicular to the bisector) are not very specific to AF
• - Pattern recognition methods, which can ensure higher quality statement, require the transmission of larger amounts of data and consequent increased technical effort in terms of storage and transmission
• - the transmission / evaluation of larger amounts of data would have to be done at regular intervals (early diagnosis)

Lorenz plots Although, therefore, in a certain way, they reflect graphic information about the Presence of AF, but these are not sufficient specified to an automatic evaluation with little effort permit.

The present invention is based on the consideration that the basic properties of the RR intervals in the presence of AF, however, should in principle make it possible to provide the appropriate starting points for the calculation of diagnostic parameters. The central characteristic of RR intervals in AF is their general irregularity. While in the sinus rhythm, with proper conduction through the AV node, RR intervals reflect the intact regulatory mechanisms of the circulatory system, and thus fluctuate in a lawful and narrow range, this is not the case with atrial fibrillation. The reason for this observation, in simplified terms, is that the irregular atrial activations during AF randomly arrive at the AV node. A transition into the ventricle occurs when the AV node is no longer refractory. The fact that the refractory time of the AV node also depends on the last routed RR interval results in a characteristic envelope towards shorter times, with the RR intervals scattering in a relatively homogeneous and irregular manner toward longer times. This specific form of irregularity appears neither in the usual sinus rhythm nor in premature ventricular beats. On the other hand, in the Lorenz plot, the geometric image of a broad scattering with envelopes in a similar form can also be observed in the presence of a large number of ventricular extrasystoles (cf. 3 ), so that from this representation a specific charac It is difficult to achieve this with simple stray dimensions that would be feasible even with the limited resources in an implant.

The method according to the invention is therefore based not on the analysis of Lorenz plots of the RR intervals, but on the evaluation of the time behavior of repeated identical RR intervals. Only RR time series are to be analyzed in a corresponding way: First of all, the different RR intervals occurring at the patient are divided according to their duration (interval classification). In a further step, the time of the last measurement of each different RR interval is stored. For each different RR interval, the number of heartbeats that it took for this RR interval value to recur is subsequently recorded. For the various possible durations of heartbeats (0, 1,..., T _j ,..., T _max ) until the repetition of an identically long RR interval, a counter C (j) is provided in each case Event (identical interval after j heartbeats found) is increased. Overall, this results in the frequency of differently long return periods in intervals of the same length after a certain measuring time. In the case of AF and sinus rhythm, this frequency distribution has a characteristic course, from which a diagnosis can be made. The differences of the two courses can be processed externally either after telemetric transmission of the few frequency values or, due to their simplicity and low data volume, also permit reshaping and evaluation in the implant so that only the transmission of the result remains (eg identification of occurring AF) as marker, single flag). The theoretical sensitivity and quality of the method according to the invention for the detection of chronic AF by means of analysis of the time behavior of the RR intervals is shown below with few assumptions regarding physiological and pathological cardiac rhythms.

With 1, ..., i, ... B _max as a sequence of binary numbers (binary numbers: "bins") for the classification of RR intervals of different lengths and T the number of heart beats until the repetition of a certain interval i with the counters C given above as probability P (T) for this return time: P (T) = C (i) / B Max (1)

Here, P (T) is normalized to the number B _{max of} all possible RR intervals specified as possible (the choice of the normalization constants is of little influence on the qualitative behavior of the probability distribution).

Furthermore, it should be taken into account that in a currently considered RR time series not all possible values (B _max ) of RR intervals occur, but only a subset k of different intervals. This value appears to be greater with AF occurring than with normal sinus rhythm (with circadian variations unaffected by these considerations, since the return times are at identical intervals). Assuming that the k different observed intervals occur stochastically with nearly equally distributed frequencies, the probability of finding one of the intervals (with the index i) can be expressed as: p (i) = 1 / k (2)

Thus, the probability of finding exactly the number T of heart beats for this interval as the return time can be stated as: P (T) = p (j ≠ i) (T-1) P (i) = [1- (1 / k)] (T-1) · (1 / k) = [1- (1 / k)] T / (k-1) (3) because the statement that the interval I recurs exactly after T heart beats is equivalent to the finding that for T-1 heartbeats this interval did not occur again, ie the probability for this "non-occurrence" just as 1 - p (i) can be expressed.

The expression in square brackets in (3) is less than 1, so the probability decreases monotonically with increasing T ≥ 1. Here is the second derivative P '' (T) = T 2 [1- (1 / k)] (T-2) + [1- (1 / k)] (T-1) (4) always positive, ie the curve of the probability distribution is convex. With the further assumption that with chronic AF the variations of the measured RR intervals fluctuate over all possible values, thus k ≈ B _max
applies: [1- (1 / k)] ≈ 1 (5) and thus P (T) ≈ 1 / B Max (6)

This probability distribution supplies the same value for each return time T, ie in graphical representation the probability course is not convex, but a flat straight line (depending on the maximum value used for B _max ).

Due to the irregular spread of the RR intervals, short return times are relatively rare in atrial fibrillation. If z. B. an interval of Width 7.8125 ms (sampling rate 128 Hz) in one step, a new beat interval will return to the original interval with little probability due to the large width k of the distribution and the missing correlation between two consecutive beats in the next beats. For sufficiently long times, however, the RR interval must return to the RR interval, ie for long return times the probability is disproportionately high.

at Sinus rhythm, on the other hand, correlates successive RR intervals (moderate). For one stationary Condition, several times during the day certainly longer time is maintained, the RR interval fluctuates multiple times, for example around an average. These oscillations inevitably result within less punches back to the original one RR interval back. The probability of the return The starting point after just a few cycles is therefore against total irregularity clearly increased, for long return times The probabilities decrease more clearly, because correspondingly long Routes without interim return occur less frequently. This is true even if the stationary conditions change or are rare.

It can therefore be assumed that the normal sinus rhythm shows an almost stationary behavior, ie it oscillates only slightly by an average value with a time frame of approximately 25 heartbeats (which corresponds to less than half a minute). Accordingly, fluctuations of the RR intervals (and return times to identical intervals) over such range are relevant and sufficient to detect anomalies (ie, T _max = 25).

5 shows for different values k the probabilities for the occurrence of different return times (up to T _max = 25) as results of the theoretical calculations according to the above derivation. Small values k (2, 5 and 10 are entered) correspond to the finding of less different RR intervals, as might be expected for a reasonably regular sine activity. The associated probability distributions of the return times show the above-derived convex course. When k = 100 different intervals occur, the expected nearly flat linear course is obtained.

In order to check the realizable significance of the method according to the invention, various publicly available data sets (physionet.org) were analyzed according to the above procedure. 6 shows the results in the illustration accordingly 5 ,

All examined records were classified as sinus rhythm, with contamination by ventricular Extrasystoles about 5% -10% is. However, the classification as sinus rhythm is for record R14149 obviously not applicable, since he is due to his Lorenz plot, its spectrum, its mean heart rate, its small number at ventricular Extrasystoles and the ECG history with high probability can be identified as permanent atrial fibrillation. The classification in the Database may be due to the fact that the software used the surface signal improperly evaluated, and adequate control by a practiced one Observers did not take place. The few detected ventricular extrasystoles were presumably by the morphology analysis of the used Software determined correctly.

The evaluation of the data by means of the method according to the invention shows for the data record R14149 in comparison to the other data sets the expected flatness for AF, in accordance with the theoretical predictions accordingly 5 , up to a return time of about 10 heartbeats and less runs in the chart of the 6 the return probability calculated from this data set is significantly flatter than for the other data sets (the range of incipient deviation is in 6 marked with a circle cutout). Thus, atrial fibrillation is detected here.

Since the other records under Sinusrhythmus despite in each case different portions of extrasystoles in 6 all show a similar course, it can be assumed that the method according to the invention against varying extrasystolic fractions is up to 10% or even more stable, since these obviously only slightly increase the return times, or reduce the probabilities for short return times by only a few percent.

The Results show that for atrial fibrillation, longer return times are preferred and short return times occur with a significantly lower probability than at sinus rhythm. This basic statistical feature is apparently very specific to chronic atrial fibrillation, and probably does not occur in any practical relevant combination of sinus rhythm and ventricular extrasystoles on. The latter would be probably only problematic if more than a third of the beating as extrasystoles would be present in combination with variable coupling constants. Such a high one Shares in extrasystoles, which in evaluation with the proposed Method pretend chronic AF could are physiologically extremely unlikely. A distinction between the two extreme states "sinus rhythm" and "chronic AF" is therefore with the method according to the invention the analysis of return probabilities in most cases automatically feasible.

For the inventive method have to for one Implementation in patient-carried portable devices, in particular Implants, computational effort and memory requirements be low. Preferably In addition, the obtained characteristic data should be transmitted daily without significantly reducing the implant life. in the Below is an embodiment of the corresponding schematically Data preparation considering the required resources outlined that meets all these criteria:

Data acquisition and calculation effort per cycle:

• - Determination the time index of the RR interval (in Implantten in front)
• - difference formation the cycle number for the last occurrence of this time index
• - provided less than 25 cycles passed: Increase the return counter C with index j of the temporal Distance in heartbeats
• - To save the new time of occurrence of the interval (for counting to return)

This flowchart is in 7 reproduced as Nassi-Shneiderman diagram.

For the illustrated sequence, only integer counters and time memories are required as variable types. For this purpose, for most situations, already three bytes of memory per value, because so can z. For example, the detectable return to a time index is greater than 65535 cycles. With these specifications, an estimate for the internal storage effort in an implant results in 384 × 3 bytes = 1152 bytes for the last return time of a RR interval (a savings option additionally results by discarding non-useful RR intervals). For a return histogram (corresponding to 5 or 6 ) already satisfy 25 × 3 bytes = 75 bytes. This reduces the daily transmission overhead to these 75 bytes for the return histogram plus 3 bytes for the total number of RR intervals, for a total of at most 78 bytes per day or 2340 bytes per month.

Further savings and thus increase The lifetime of an implant is quite possible because of storage and transmission requirements often to 2 bytes per counter is reducible because larger values as 65535 in the return histogram extremely unlikely are. With a reduction of transmission to 53 bytes daily (= 1590 bytes per month), probably the lifetime shortening by again reduced power consumption a few months or less shorter be.

The external implementation all further evaluations are with daily transmission of the return histogram easily possible and can z. For example setting thresholds, storing the time history and study for changes like sudden jumps be used.

In an expanded performance form The invention relates to data processing means in the implant integrated, for example as application-specific circuits (ASICs) or microprocessors (μP). So then the individual evaluations are also in the implant itself possible. An occurrence of AF could then be detected by the implant itself and corresponding information for further diagnosis become.

The Invention allows detection in patients with single-chamber systems of atrial fibrillation. It can be done by a telemetric transmission a quick diagnosis and early on onset therapy of AF can be achieved. The externally transmitted The amount of data is very low and thus allows a problem-free review of the Patent status, e.g. In 24-hour rhythm. An elaborate and error-prone evaluation deleted from morphological criteria, there exclusively on the readily available RR interval used becomes. The AF detection according to the invention also works with normal ventricular rhythm and without Existence of other tachycardia criteria.

A Apparatus for implementing the method according to the invention requires no special expenses, making it also suitable for single chamber equipment is. The implants to be performed in the implant Calculations involve a few simple counting operations per cardiac cycle, the algorithm to be used is limited to storage, manipulation and transmission integer values. Data storage in the implant requires a few hundred bytes. Such functions can already be very simple DV components (eg ASIC or μP). Of the Energy consumption is correspondingly low.

The Invention allows safe and easy detection of atrial fibrillation, where, for the first time, single-chamber systems could also be used.

Claims (8)

Method for evaluating ECG signals, wherein detected by means of electrodes heartbeats as a sequence of R peaks and z. B. are recorded as RR time series, characterized in that the RR intervals occurring after a predetermined number (B _max ) of different time intervals are classified and the respective times fed up to repetition of an identical interval chert and analyzed. A method according to claim 1, characterized in that different possible repetition times (T ₁ ... T _j ... T _max ) until the re-occurrence of an identically long RR interval are given and each predetermined repetition each time a counter C (j) such is assigned to the counters, the number of the respective events, ie finding an identical interval after repetition time j, is stored and wherein the distribution of C (j) is analyzed over the different repetition times. Method according to claim 2, characterized in that that the course of the distribution C (j) is analyzed to that an approximate equal distribution the C (j) over the different repetition times j as markers for atrial fibrillation (AF) is set. Method according to one of Claims 2 or 3, characterized in that the different possible repetition times (T ₁ ... T _j ... T _max ) are specified as the number of heartbeats occurring up to the repetition. Method according to one of claims 1 to 4, characterized that the number k of the occurring different RR intervals detected and used for the evaluation. Device for carrying out the method according to one of claims 1 to 5, wherein the device is designed as a portable ECG device or as an implant with means for detecting heartbeats as a sequence of R-peaks, characterized in that means for classifying occurring RR intervals are present after a predetermined number (B _max ) of different time intervals, as well as means for detecting the respective times until the repetition of an identical interval, as well as means for storing and transmitting a corresponding data set. Device according to claim 6, characterized in that means for storing the frequency of repetition times that occur available. Device according to claim 7, characterized in that that means (ASIC, μP) for the analysis of the frequency distribution particular to equal distribution, as well as means of transmitting the frequency distribution and / or the results of the analysis.