EP1570378A1

EP1570378A1 - Method for evaluating a sequence of discrete readings

Info

Publication number: EP1570378A1
Application number: EP02808173A
Authority: EP
Inventors: Georg Schmidt; Axel Bauer
Original assignee: Individual
Current assignee: Schmidt Georg
Priority date: 2002-11-27
Filing date: 2002-11-27
Publication date: 2005-09-07
Also published as: WO2004049190A2; US20060064285A1; US7200528B2

Description

Method for evaluating a sequence of discrete

readings

The invention relates to a method for evaluating a sequence of discrete measured values according to the preamble of claim 1.

The investigation of followers of discrete measured values is common in many areas of technology. So z. B. for functional testing of machines with rotating or reciprocating machine parts vibration patterns evaluated using the Fourier analysis, for example, in order to

To determine the frequencies contained in the vibration pattern. From the frequencies determined, otherwise undetectable malfunctions can be inferred, e.g. B. on beginning damage to bearings etc.

The invention has for its object to provide a method for evaluating a sequence of discrete measured values, such. B. to specify measured values based on vibration patterns, repetitive processes or processes that change over time, with which parameters can be determined which define characteristic properties, including hidden ones, of the measured value sequence.

This object is according to the invention by Features of claim 1 solved.

Accordingly, the method for evaluating a sequence of discrete measured values has the following features:

an attribute is assigned to each measured value, this attribute having the same property for all measured values;

from the measured values thus assigned with attributes, those are selected whose attributes each fulfill a certain common criterion and are each referred to as a certain measured value or event;

a chain of a certain length is selected from the sequence of discrete measured values;

partial chains are formed from this selected chain, each of which has a selected event as the center, the remaining measured values of the partial chains being arranged in the original sequence on both sides of the central event;

the measured values of the individual partial chains are assigned to one another according to their position in the respective partial chain, namely the measured values of the central events are assigned to one another and the measured values that are in the same place around the respective central event;

all assigned measurement values for the same place within the selected partial chains are averaged and the sequence of the measurement values averaged in this way in the range specified by the chain of measured values

Order recorded.

According to the invention, all measured values of the signal are thus assigned a certain attribute in a first step, which characterizes the measured values in any desired property. In principle, any kind of attribute is conceivable, in a simple case e.g. the attribute can consist of the difference to the previous measured value. However, the attribute does not necessarily have to be derivable from the signal to be analyzed. It is also possible to assign an attribute from a synchronously recorded second signal to the measured values of the signal to be analyzed, e.g. the measured values of an EKG signal with attributes from a blood pressure signal recorded at the same time.

At least some measured values are selected from the measured values thus assigned attributes, the attribute of which fulfills a certain criterion. If, for example, all measured values were assigned the attribute of the difference to the previous measured value, the selection criterion could be that the attribute is negative, ie the measured value is larger than al? the previous one is. The selected measured values are to be called “events” in the following. A chain of measured values is now created for each event, the central element of which is the event itself, surrounded by the measured values which are arranged in the output signal before or after the event Chains formed in this way for each event, the lengths of which can be freely defined, are arranged “one below the other” in a table, each row of the table corresponding to a chain of measured values. It is important that the Measured values of the same positions with respect to the

Events are one below the other. Is that true

Criterion towards a relatively large part of the measured values depends on the selected one

Chain length, measured values of the output signal in the

Represent the table multiple times, i.e. the table contains redundant information according to the methodology. In a next step, the averages of each column in the table are formed, i.e. the measured values of the same position with respect to the selected one

Measured values are averaged.

In this way, an output signal of any length is transformed into a relatively short sequence of averaged measured values, hereinafter referred to as "Schmidt-Bauer transformation (SBT)". The SBT characterizes the output signal of any length, depending on the selected attribute and the selected criterion, which one The SBT can be evaluated in different ways, for example in the time or frequency domain.

It has been found that the evaluation proposed by the invention makes it possible to determine a large number of parameters which cannot be seen from the sequence of the measured values alone.

So z. B. from a vibration pattern such as a machine with rotating or reciprocating machine parts, the frequencies underlying the vibration pattern are determined, similar to a Fourier analysis. B., "Interference frequencies" are due to damage to the machine or its parts indicate. It has been shown that the method according to the invention also determines frequencies which are used in the conventional fast Fourier

Transformation remain undetected.

The method according to the invention can be used particularly advantageously in medical technology, e.g. B. for the evaluation of long-term ECGs, long-term blood pressure measurements etc. With regard to the evaluation of long-term ECGs in infarct patients, the method can be used as follows. The sequence of beat-to-beat intervals of the long-term ECG serves as the output signal, which is the basis of most evaluation methods of long-term ECGs. An attribute is first assigned to each measured value. The comparison to the previous measured value should serve as an attribute, more precisely the quotient between the measured value itself and the previous measured value.

Is a measured value e.g. 3% shorter than the previous one, the attribute "0.97" would be assigned to this measured value. If it is 3% longer, the quotient is 1.03. The criterion for the selection of a measured value could be a value between 1 and 1.05 According to this criterion, all measured values would be selected as an event, which are larger, but not larger than 5% of the previous measured value .. In an average long-term ECG approximately 30% of all measured values correspond to this criterion.

The SBT corresponds to the average course of the measured values before, during and after the event (here: rise in the measured values in the output signal). The SBT can be quantified in many ways: For the Estimating future risk in patients

Heart attack has turned out to be cheap, the central one

Sum the value of the SBT with the subsequent value and subtract the sum of the two previous measured values from this sum. This so-called

EDC2 value can be seen as a measure of the central increase and allows a statement regarding the patient's chance of survival. The bigger this

EDC2 value, the larger the

Patient's chance of survival. However, the SBT only quantifies the EDC2 value in immediate

Surroundings around the center, but parameters are also conceivable that capture longer-term changes in the SBT.

In the evaluation just described, an attribute was assigned to the measured values of the original signal, which establishes the relation to the immediately preceding measured value, whereby the SBT emphasizes above all the shorter frequencies of the output signal. An emphasis on longer frequencies can be achieved by assigning attributes to the measured values that compare the measured values over a longer period of time, e.g. compare the sum of the values of a certain period after the measured value with the sum of the values of a certain period before the measured value.

It is also possible to assign the measured values of a time series (eg those of the ECG) with an attribute from a time series recorded at the same time (eg increase in blood pressure from a blood pressure curve). If both time series are linked (in this example by the baroreflex: increase in blood pressure causes pulse slowdown, causes blood pressure drop

Pulse acceleration), the SBT filters significantly

Vibrations from the EKG signal. If not

Case, there are no Scr-oscillations. The SBT is therefore also suitable for the detection or exclusion of

Couplings of different signals.

Further embodiments of the invention emerge from the subclaims.

The invention is explained in more detail in exemplary embodiments with reference to the drawing. In this represent:

Figure 1 with lines a to f is a schematic representation of the method for evaluating a sequence of discrete measured values according to the invention;

FIG. 2 shows a so-called tachogram in which the time intervals between two successive heartbeats are plotted for a patient over a period of 24 hours;

FIG. 3 shows the frequency spectrum of the tachogram in FIG. 2, which was formed using a simple Fourier transformation;

Figure 4 shows the Schmidt-Bauer transformation of the tachogram in Figure 2 according to the invention;

FIG. 5 shows a detail from FIG. 4, which illustrates the calculation of the parameter EDC2;

FIG. 6 shows the frequency spectrum of the Schmidt-Bauer transformation in FIG. 4 7 shows the Schmidt-Bauer transformation similar to FIG. 4 of a risk-prone infarct patient;

FIG. 8 shows the frequency spectrum of the Schmidt-Bauer transformation according to FIG. 7;

FIG. 9 shows the distributions of the EDC2 parameter for infarct patients who have survived for a longer period and for infarct patients who have died within a certain period of time;

FIG. 10 shows the amplitude of an oscillation over time, in this case a rotational acceleration of a truck wheel with unbalance;

FIG. 11 shows a simple Fourier transformation of the oscillation according to FIG. 10;

FIG. 12 shows the Schmidt-Bauer transformation of the vibration according to FIG. 10 on the basis of an evaluation using a method according to the invention; and

FIG. 13 shows a Fourier transformation of the Schmidt-Bauer transformation shown in FIG. 12.

The general principle of the invention is shown in FIG. 1 with lines a), b), c), d), e) and f).

The sequence of the measured values to be analyzed is plotted in line a).

According to line b) a certain amount is released for each measured value definable attribute assigned, eg the difference to the previous measured value.

According to line c), a criterion is defined which is applied to the attributes assigned to the measured values. In this example, the criterion is met (K = 1) if the attribute is positive (A> 0). Measured values whose attributes meet the defined criteria are defined as events.

According to line d), the events, together with the previous and following measured values, are entered as event chains in a table, in such a way that measured values of the same position are related to one another for the event.

According to line e), the Schmidt-Bauer transformation is determined by averaging the event chains. In this way, the events themselves and the measured values of the same position with respect to the event are averaged.

In Figure 2 is a tachogram, i.e. H. the sequence of the time intervals between two successive heartbeats (= RR intervals or RRI), in this case approximately 80,000 time intervals, is plotted against their number on the EKG. This tachogram is from a patient after a myocardial infarction who has survived it for over two years.

FIG. 3 shows the frequency spectrum of the tachogram shown in FIG. 2, which can be calculated using a simple Fourier transformation as a function of Amplitude was determined depending on the frequency.

FIG. 4 shows the Schmidt-Bauer transformation of the tachogram in FIG. 2 according to the invention. The quotient between the measured value itself and the previous measured value was used as an attribute, and a value between 1 and 1.5 was used as the criterion. The chain length of the Schmidt-Bauer transformation was set to 200 here, i.e. 100 values before and 100 values after the central event. The values of the transformation fluctuate around the value 1000 ms, whereby considerable dynamics can be determined around the center at the value zero. Fast modulations of the beat-to-beat intervals can be distinguished from slower ones.

FIG. 5 shows a section around the center of the Schmidt-Bauer transformation according to FIG. 4 and illustrates the calculation of the parameter EDC2, which is used for risk stratification. The values RRI (-2), RRI (-l) and RRI (+1) are entered around the central event RRI (0). From these values, a character, static value EDC2 can be calculated, namely

EDC2 = (RRI (0) + RRI (+ l)) - (RRI (-l) + RRI (-2)

With the values of the curves shown, a value of EDC2 = 32 ms is calculated. FIG. 6 shows a simple Fourier transformation of the Schmidt-Bauer transformation according to FIG. 4. You can see clear swings at around 0.003Hz, at 0.03Hz and at 0.3Hz, which in the simple Fourier transform of

Output signal are not recognizable. These three

It is known that frequencies correspond to various physiological mechanisms of the

Heart rate regulation. For example, the

Rash at 0.3Hz of breath modulation

Heartbeat. An additional one is also recognizable

Deflection between 0.003Hz and 0.03Hz, which is not yet known.

FIG. 7 shows a Schmidt-Bauer transformation of a tachogram of a patient who suddenly died within two years of a heart attack. You can see that the dynamic around the central event is clearly limited. The EDC2 value is significantly lower at 6 ms.

FIG. 8 shows the Fourier transformation of the Schmidt-Bauer transformation according to FIG. 7 (the Fourier transformation according to FIG. 6 is shown as a dotted line for comparison). The deflection at 0.3 Hz that can be seen in FIG. 6 is not present, but there is a deflection at approximately 0.75 Hz. This corresponds to the modulation of the heartbeat by the pressure receptors of the carotid artery, the so-called baroreceptors.

FIG. 9 shows the differences in the distribution of the EDC2 values calculated for over 1000 patients for patients who had a monitoring interval of two Have survived for years (group 0) and for patients who have died during this period (group

1) . It can be seen that the EDC2 values of the patients who died are significantly lower than the EDC2 values of the patients who survived this period. The significant predictive power of EDC2 values has been tested and proven on over 4000 infarct patients.

FIG. 10 shows the time behavior of the amplitude of an oscillation, in this case the oscillation in connection with the rotational acceleration of a truck wheel with an imbalance. The amplitude as a function of time is plotted in this figure.

The Fourier transformation of the original vibration corresponding to FIG. 8 is plotted in FIG.

FIG. 12 shows the Schmidt-Bauer transformation of the vibration according to FIG. 10. The difference to the previous measured value was used as an attribute and a value above zero was used as the criterion.

FIG. 13 shows the Fourier transformation of the Schmidt-Bauer transformation according to FIG. 10. It can be seen that this representation essentially corresponds to the diagram according to FIG. 11.

A decisive advantage is that the transformation represents a relatively short signal, which, however, does not lose any information in comparison to a long output signal.

Claims

claims

1. Method for evaluating a sequence of discrete measured values, with the following features:

all assigned measured values for the same Space within the selected sub-chains are averaged and the sequence of the averages so

Measured values are recorded in the order specified by the chain of measured values.

2. The method according to claim 1, characterized in that the attribute for each measured value is derived from other measured values of the measured value sequence and / or from measured values of a second measured value sequence recorded synchronously with this first measured value sequence.

3. The method according to claim 1 or 2, characterized in that the attribute is derived from the respective previous measured value.

4. The method according to any one of the preceding claims, characterized in that the difference to the previous measured value is selected as the attribute for each measured value.

5. The method according to claim 4, characterized in that for the attribute the criterion "measured value lies in a certain range in relation to the previous measured value" is recorded.

6. The method according to any one of the preceding claims, characterized in that defined measured value groups or functions of the measured values or measured value groups are selected instead of individual measured values.

7. The method according to claim 6, characterized in that the criterion "larger" or "smaller than that." previous measurement "is recorded.

8. The method according to claim 7, characterized in that the criterion "recorded in a certain area larger" or "smaller in a certain area than the previous measured value" is recorded.

9. The method according to any one of the preceding claims, characterized in that the percentage deviation is recorded as a criterion for successive measured values.

10. The method according to any one of the preceding claims, characterized in that the evaluation of the sequence of discrete measured values takes place in the time or frequency domain.

11. The method according to any one of the preceding claims, characterized in that the chain of measured values has 200 or more measured values.

12. The method according to any one of the preceding claims, characterized in that the number of selected chains is more than 4 and is preferably in the tens range.

13. The method according to any one of the preceding claims, characterized in that each measured value represents a physiological variable.

14. The method according to claim 13, characterized in that the physiological size the time interval between two heartbeats.

15. The method according to claim, characterized in that the physiological size is the blood pressure.

16. The method according to any one of claims 1 to 12, characterized in that oscillation patterns are evaluated with the method, the oscillation pattern being sampled at a high frequency and the evaluated measured values being subjected to a Fourier transformation.