EP1097439B1 - Method and device for detecting drifts, jumps and/or outliers of measurement values - Google Patents

Abstract

A method for detection an alarm state of measurement signal values received by means for detecting measurement values, wherein an alarm state is triggered when for a currently received measurement signal value at least one pre-set limit value is exceeded, has a faster recognition of an alarm situation and at the same time a lower rate of false alarms when in a first step for measurement signal values which are subsequent in time their position parameter (2) and a corresponding deviation parameter (3) of the measurement signal values from the position parameter is calculated in an adjustable time window, wherein in a second step each further subsequent measurement signal value is compared to the position parameter (2) an weighted with the deviation parameter (3) in order to obtain a respective evaluation quantity, and wherein in a third step an outlier state (6) is detected when the evaluation quantity exceeds an adjustable outlier parameter, whereas an alarm state (8) indicating the presence of a significant drift or jump of the measurement signal values is detected when the evaluation quantity exceeds an adjustable alarm parameter.

Description

The invention relates to a method for detecting an alarm condition or for Detection of drifts, jumps and / or outsiders by means of measured value acquisition means received Meßsignafwerten, with an alarm condition triggered is, if for a currently received Meßsignalwert or for one off Measured values derived value a predetermined limit or predetermined Interval limits are exceeded or will be. As examples of the utmost numerous applications of the method according to the invention are in particular in the field of medicine perioperative monitoring, monitoring of vital signs in intensive care units, sleep monitoring, CTG (Cardio-tocography), and in other areas, fire and smoke alarm systems, acoustic monitoring systems, such. B. babyphone called.

1. Geringe Fehlalarmrate, um den unerwünschten Effekt der Gewöhnung an die Alarmsituation zu vermeiden und um der Neigung zum Desaktivieren des oft als störend empfundenen Alarms entgegenzuwirken.

2. Kurze Verzögerungszeiten zwischen Anbahnung einer krititschen Situation und der Auslösung des Alarms, um einen unter Umständen lebenswichtigen Zeitvorsprung für therapeutische Eingriffe zu gewährleisten.

3. Hohes Maß an Adaptionsvermögen, um zu vermeiden, daß allzuviele Parameter von Hand voreingestellt und während der Behandlung nachgestellt werden müssen, und damit von der eigentlichen Überwachungsaufgabe ablenken. Insbesondere sollen mehrere mit gewissem Zeitabstand aufeinanderfolgende Alarmsitutationen erkannt werden können.

4. Hohes Maß an Aussagekraft der einstellbaren Parameter, um zu gewährleisten, daß das Alarmsystem auch leicht und fehlerfrei bedienbar ist.

5. Größtmögliche Einfachheit und damit größtmögliche Rechengeschwindigkeit, um aufwendige Rechnungen zu vermeiden, die nur mit teuren Prozessoren und Speicherelementen durchführbar wären, und um etwaige Rechenzeitbeschränkungen zu umgehen.

6. Große Aussagekraft, um differenzierte Reaktionen zu ermöglichen.

7. Integrierte Erkennung von Ausreißern, um eine Differenzierung zwischen lebensbedrohlichen Zuständen, Gerät- und Zuleitungsversagen und Fehlmessungen zu ermöglichen.

8. Klare Entscheidungsregeln, um Exportierbarkeit zu gewährleisten und eine retrospektive Analyse und Parameterkorrektur zu ermöglichen.

Intensive care monitor alarm systems, typically representing and analyzing cardiovascular parameters (ECG, blood pressure), oxygen saturation (SpO2), gas exchange and metabolic parameters, as well as EEG and EMG online, are designed to alert the attending physician or nurse to potentially life-threatening conditions for the monitored patient to steer. An ideal alarm system would be characterized by the following features, all of which are not optimally realized in the prior art:

1. Low false alarm rate, to avoid the unwanted effect of habituation to the alarm situation and to counteract the tendency to deactivate the often annoying perceived alarm.

2. Short delay times between the initiation of a critical situation and the triggering of the alarm, in order to ensure a possibly vital time lead for therapeutic interventions.

3. High level of adaptability to avoid too many parameters having to be pre-set by hand and readjusted during the treatment, thus distracting from the actual monitoring task. In particular, several consecutive alarm situations should be detected with a certain interval.

4. High degree of predictive value of the adjustable parameters to ensure that the alarm system is also easy and error-free to use.

5. Greatest possible simplicity and thus the highest possible computing speed to avoid complex calculations that would be feasible only with expensive processors and memory elements, and to circumvent any computing time limitations.

6. Great meaningfulness to allow differentiated reactions.

7. Integrated outlier detection to allow differentiation between life-threatening conditions, device and lead failures and erroneous measurements.

8. Clear decision rules to ensure exportability and allow for retrospective analysis and parameter correction.

Current alarm systems in intensive care have a false alarm rate of 70% to 99.5%, depending on the monitored physiological Size. The high false alarm rate leads to desensitisation of the monitoring staff and for frequent manual alarm deactivation. The well-known Alarm systems are triggered when the size to be monitored is preset exceeds upper or lower limits. Such alarm systems will be referred to as threshold alarm systems. To lower the false alarm rate The upper limit is rather high and the lower limit rather low which, however, inevitably becomes larger in alarming situations Time delays leads. In addition, such an all-or-nothing system corresponds not the ISO standard, which has a tiered alerting system with different Suggesting menses.

In the known threshold alarm system is for a fluctuating signal a given upper and lower threshold, where, when the signal is off an alarm is triggered which moves the interval defined by the thresholds becomes. The threshold alarm has the following disadvantages. He is unstable towards Outliers. He is not adaptive, d. H. Borders have to be set by hand and, especially with a signal with a drift, z. B. caused by a temporal Change in detector sensitivity, permanently adjusted. Become the limits of the threshold alarm set too far, it comes to long Delay times until an alarm is detected. If the limits are too narrow on the other hand, false alarms often occur. Therefore, in practice, a so-called "borderline" or an option, such as "all alarms off for two Furthermore, the threshold alarm system is not in the case suitable that a plurality of signals are monitored by an alarm system got to.

The prior art is referred to the patent DE 35 23 232 C2. From this document is a fire alarm system for detecting and delivering a a change in a physical appearance of the environmental conditions corresponding analog value known. There are a scanning device for sampling an analog output from a detection section Detection signal within a certain period of time, a data processing device for forming an average value from the sampling data, as well as a Memory device in which these sample data can be stored and an alarm device, the presence of a fire after evaluation of the mean indicates provided. Characteristically, the data processing device is formed such that the sampling data sequentially in the memory device be inscribed, and from a certain number of the last saved Sampling data is continuously formed a moving average, each the oldest in the sequence sample data storage value by the newest is replaced.

Furthermore, from DE 31 27 324 A1 discloses a method and an arrangement for increasing the sensitivity and noise immunity in a hazardous, in particular Fire alarm system, known.

From both publications mentioned above is not known in particular a Calculate scattering parameters from the successive measured values, so that the alarm triggering method is adaptive, and thus capable of learning. Therefore, the above methods are not capable of, for example, a temporal Adjust the change in detector sensitivity.

Finally, DE 44 17 574 C2 relates to a patient alarm detection under Using a target mode. In this procedure, at an intended Changing a physiological parameter of a patient dynamic Defines limits and then generates an alarm when the measured parameter values outside the dynamic limits. This document discloses Therefore, only a variant of the known threshold alarm.

The object of the present invention is therefore to overcome the disadvantages of To avoid prior art, and in particular a method of the beginning so-called type such that over the prior art an "alarm situation" detected faster and with a lower false alarm rate becomes.

The object is achieved in procedural terms by the features of claim 1.

Thus, in the method according to the invention, a distinction is made between two phases, wherein in a first phase a time window is provided in which evaluated the characteristic course of the measured signal values detected therein with the statistical average and the fluctuation range of the detected Measured signal values around this mean value is determined. In the second phase of the invention The method becomes the currently received measured signal values with the mean and the fluctuation width representing scatter compared with the determined evaluation size is a measure of the existence represents a significant drift. By putting into this evaluation variable thus obtained the temporal evolution of the measured signal values acquired in the time window the overall result is a higher degree of reliability in the recognition of alarm conditions over prior art methods, so that can therefore achieve a lower false alarm rate. This is based in particular on the automatic tracking of the interval limits. Due to the of the method according to the invention, leading to adulterations and / or false alarms, outlier states and alarm conditions becomes a differentiation in intensive care applications between life-threatening conditions on the one hand and incorrect measurements leading device or supply failure on the other hand allows what to a further reduction of the false alarm rate leads.

An advantage of the method according to the invention is that an on-line Detection of outliers is provided. Femer is advantageous that the inventive Method is adaptive, d. H. for example, only physiological Limits must be preset. Furthermore, according to the invention drifting and / or jumps are detected automatically. Finally, the inventive Process only a short delay time.

In order to achieve a high computing speed, the evaluation size becomes by subtraction between the measured signal value and the calculated mean value with subsequent normalization of the difference. Here is the Weighting of the valuation variable by a division formation from the normalized Difference between the measured signal value and the mean value with the calculated value Scattering made.

According to an advantageous embodiment of the method according to the invention is an outlier state is detected when the one weighted with the calculated scatter normalized difference between measured signal value and mean value of the set Outlier parameter exceeds. On the other hand, an alarm condition is detected, if the weighted normalized difference between Measured signal value and mean exceeds the set alarm parameter.

To measurement errors, for example, by equipment failure or metrological Artifacts come about, eliminate, will occur when a Outlier state the corresponding measurement signal value by the current, in the calculated window, the mean value replaced and the next following Measured signal value edited.

Alternatively, a different substitution may be made, which is particularly preferred for statistical reasons. It can, for example a noise can be added or another imputation performed become. In this case, the outlier value can be determined in particular by an average value plus added random number, which derives from a probability distribution, be replaced. Finally, such a falsifying or falsified Measured value for the further calculation also simply be ignored.

It has proved to be expedient if the mean of the successive Measurement signal values formed from the summation of the individual measured signal values is, where the number of summation steps by the width of the time window is determined. The standard deviation is used as the scatter, wherein the number of summation steps is determined by the width of the time window becomes.

A computationally advantageous embodiment of the invention Method is that by means of a time delay a positioning of the time window is made, even small gradients to be able to recognize in the time course of the detected measured variable, so that also long-term drifts by a correspondingly far away, delayed Detect window (delayed moving window). Similarly, short-term Drifting at a correspondingly close, delayed window (delayed moving window) are recognized.

To distinguish between occurring outlier conditions and alarm conditions To facilitate, the outlier parameter becomes a higher value set as the alarm parameter.

It has proved to be particularly useful when the width of the time window preferably set to 10 temporally successive measured signal values and the outlier parameter is set to 6 and the alarm parameter is set to 3 becomes.

In device technical terms, the above object is achieved by the Characteristics of claim 14 solved.

By interaction of the individual components are thus in accordance with a thus obtained evaluation size outlier states and alarm conditions distinguishable from each other, so that therefore the false alarm rate against Significantly reduce the method of the prior art.

Fig. 1

ein Flußdiagramm mit den wesentlichen Prozeßschritten des erfindungsgemäßen Verfahrens;

Fig. 2

ein Meßwertspektrum der zeitlichen Entwicklung einer physiologischen Meßgröße;

Fig. 3a

eine stark schematische Darstellung einer Drift;

Fig. 3b

eine stark schematische Darstellung eines Sprungs; und

Fig. 3c

eine stark schematische Darstellung eines Ausreißers.

With reference to the accompanying drawings, an embodiment of the present invention will be explained below. In partial schematic views show:

Fig. 1

a flowchart with the essential process steps of the method according to the invention;

Fig. 2

a measured value spectrum of the temporal development of a physiological measurand;

Fig. 3a

a highly schematic representation of a drift;

Fig. 3b

a highly schematic representation of a jump; and

Fig. 3c

a highly schematic representation of an outlier.

The method according to the invention, which preferably implements as a software program is, with its essential process steps in one in total illustrated with 10 flowchart in Fig. 1 illustrates. During an initialization phase 1, a time window is provided in which on a length of i time sequential steps for those detected in the time window Measurement signal values an average value 2 and the associated scattering 3 of the measured signal values to calculate this mean. The mean, however, will be not calculated from a series of the immediately preceding measured values, but from a time window of width ω in the past with the selectable one Time delay d. The lower summation limit for the determination of the mean value thus results from the subtraction n-d-ω, where n is the number of executed Time steps, d denotes the time delay and ω the window width. In contrast, the upper summation limit results from the subtraction n-d, so that the summation index i runs from n-d-ω to n-d. The same summation limits apply to the determination of the dispersion 3.

In the actual process phase, an incrementation is performed in a process step 4. In a further process step 5, the detected in a given time step _n measurement signal value Y is compared with the calculated mean value in the initialization phase by a difference formation carried out, and this difference is provided with a normalization amount. In order to take into account also the variance in the comparison, the absolute difference is weighted with the dispersion by including the dispersion as a divisor. The evaluation variable thus obtained serves as a measure in the detection of occurring outlier states in this process step 4. Namely, if the evaluation variable obtained for the currently detected Meßsignalwert greater than a preset outlier parameter o (o> 0), the query results in this process step 4 that an outlier condition 6 is present. The outlier condition can be ignored for the following calculation or replaced with a "reasonable" value. In particular, imputation methods are suitable for this purpose. In this case, the sequence program returns to the increment instruction 4.

If the query in block 5 results in a negative result, the query block will contain 7 determines whether the data obtained for the currently detected Meßsignalwert Evaluation size is greater than a preset alarm parameter a. At a positively determined result is an alarm condition 8 before. In the embodiment In this case, a return to the initialization phase is made while if there is a negative result, a return to the increment instruction is made. As a boundary condition in the distinction between Outlier states and alarm conditions becomes a higher value for the outlier parameter Value assigned as the alarm parameter. By the distinction between Outlier states and alarm conditions will be differentiated between significant states and false measurements, with erroneous measurements may be caused by supply line failure or metrological artifacts.

The detection and elimination of such incorrect measurements thus leads to a Reduction of false alarms.

Fig. 2 shows the time course of a physiological measured variable. The abscissa axis serves as the time axis τ _CP , while the ordinate axis reproduces the amplitude of the measured signal.

Fig. 3a shows a highly schematic representation of a drift. Fig. 3b shows a strong schematic representation of a jump. Fig. 3c shows a highly schematic Representation of an outlier. It is the temporal dependence of one measured signal.

In summary, it is thus characteristic of the method according to the invention: The internal parameters of the algorithm are the window width ω (ω> 0), the delay d (d> 0), the initialization length i (i> ω + d), the outlier parameter o (o> 0) and the alarm parameter a (a> 0).

After an initialization phase of the length of i time steps, the respective newly measured value with one estimated from the previous measured values Mean and associated dispersion (the empirical standard deviation) so far the algorithm is a natural generalization of the normal threshold alarm, where the mean and spread known be presupposed. However, the average is not from a series of calculated directly preceding measured values, but from a time window the width ω in the past, with the selectable time delay d. These Type of calculation bypasses the problem that the estimation of the mean and the spread used measurements have already begun to drift and thus contribute to a considerable bias that can go so far that a sufficiently slow drift is not recognized at all. Rather, it is about the freely selectable delay d given the possibility of the limit angle of those To choose a slope that you just want to recognize. Naturally, d the smaller the slope, the larger the chosen. Each newly measured Value is estimated with the current method according to the invention Mean value compared as follows: is the measured value more than the product from selectable outlier factor and scatter from estimated mean removed, it is classified as an outlier and is subject to further calculations the current mean (plus a random number with expected zero and Scatter according to the estimated spread). If this is not the case Case, but the measured value is more than the product of (selectable) alarm factor a and scatter away from the estimated mean, it is reported that There is a significant drift, depending on the direction of the deviation a drift up or down. In all other cases, no message is output. Then the next time step is processed. It is selectable, whether should be reinitialized after an alarm has been issued, under certain circumstances with another selectable time delay, or without a new initialization should be further calculated. The window width ω influences the fluctuations the estimated mean value - the fluctuations decrease proportionally to the root of ω.

For many purposes, the following values prove to be favorable starting values can then be optimized: window width ω> 10, outlier parameter o = 6 and the alarm parameter a = 3. The calculated information Outliers Yes / No, Alarm for drift up / down, respectively, no significant drift can either directly on the screen or acoustically via agreed tone sequences be issued, or at the entrance to an intelligent alarm system.

The invention has been described in more detail above with reference to a preferred embodiment explained. However, it will be apparent to one skilled in the art that various Modifications and modifications can be made without departing from the Deviate from the underlying idea of the invention. In particular, it should be noted that in the present description under the term "positional parameter" in particular a mean, median, or similar and under the term "scattering parameter" a standard deviation, quantile, or similar is understood.

Claims (15)

1. Method for detecting an alarm state of measurement signal values received by measurement value sampling means wherein an alarm state is triggered when for a currently received measurement signal value at least one pre-determined limit value is exceeded, characterized in that in a first step for measurement signal values subsequent in time a position parameter (2), which is a mean value, median or similar, thereof and a corresponding deviation parameter (3), which is a standard deviation, quantile or similar, of the measurement signal values from the position parameter is calculated in an adjustable time window, in that in a second step each further subsequent measurement signal value is compared to the position parameter (2) and weighted with the deviation parameter (3) in order to obtain a respective evaluation quantity, and in that in a third step an outlier state (6) is detected when the evaluation quantity exceeds an adjustable outlier parameter, whereas an alarm state (8) indicating the presence of a significant drift or jump of the measurement signal values is detected when the evaluation quantity exceeds an adjustable alarm parameter.
2. Method according to claim 1, characterized in that the evaluation quantity is calculated by taking the difference between the measurement signal value and the calculated position parameter (2) with a subsequent normalization of the difference.
3. Method according to claim 2, characterized in that the weighting of the evaluation quantity is provided by a division of the normalized difference between the measurement signal value and the position parameter (2) and the calculated deviation parameter (3).
4. Method according to claim 3, characterized in that an outlier state (6) is detected when the normalized difference which has been weighted with the calculated deviation parameter (3) between the measurement signal value and the position parameter (2) exceeds the set outlier parameter.
5. Method according to claim 3 or 4, characterized in that an alarm state (8) is detected if the normalized difference weighted with the calculated deviation parameter (3) between the measurement signal value and the position parameter (2) exceeds the set alarm parameter.
6. Method according to any of claims 1 to 5, characterized in that when an outlier state (6) is present the corresponding measurement signal value is replaced by a substitute value and the next subsequent measurement signal value is processed.
7. Method according to claim 6, characterized in that the substitute value is the current position parameter (2).
8. Method according to any of claims 1 to 5, characterized in that when an outlier state (6) is present the corresponding measurement signal value is left out for the further calculation and the next subsequent measurement signal value is processed.
9. Method according to any of claims 1 to 8, characterized in that the position parameter of the subsequent measurement signal values is formed by a summation of the single measurement signal values, wherein the number of summation steps is determined by the width of the time window.
10. Method according to any of claims 1 to 9, characterized in that as a deviation parameter (3) the standard deviation is used, wherein the number of the summation steps is determined by the width of the time window.
11. Method according to any of claims 1 to 9, characterized in that a variable positioning of the time window is provided using a time delay.
12. Method according to any of claims 1 to 11, characterized in that the outlier parameter is set to a higher value compared to the alarm parameter.
13. Method according to any of claims 9 to 12, characterized in that the width of the time window is preferably set to be ten measurement signal values subsequent in time and that the outlier parameter is set to six and the alarm parameter is set to three.
14. Apparatus for detecting an alarm state of measurement signal values received by measurement value sampling means wherein an alarm state is triggered when for a currently received measurement signal value at least one pre-determined limit value is exceeded, characterized in that the apparatus comprises calculation means to calculate in a first step in an adjustable time window for measurement signal values subsequent in time a position parameter (2), which is a mean value, median or similar, thereof and a corresponding deviation parameter (3), which is a standard deviation, quantile or similar, of the measurement signal values from the position parameter, wherein in a second step each further subsequent measurement signal value is compared to the position parameter (2) and weighted with the deviation parameter (3) in order to obtain a respective evaluation quantity, and in that the apparatus further comprises detection means which detect in a third step an outlier state (6) when the evaluation quantity exceeds an adjustable outlier parameter, whereas the detection means detect an alarm state (8) indicating the presence of a significant drift or jump of the measurement signal values when the evaluation quantity exceeds an adjustable alarm parameter.
15. Apparatus according to claim 14 comprising a measurement value sampling device for receiving measurement value signals and a measurement value transfer device for converting and processing the received measurement value signals, as well as an alarm device triggerable when at least one limit value is exceeded, characterized in that a storage device is provided for sampling the measurement signal values in a time window adjustable in width and time delay, in that calculation means are provided for calculating the position parameters (2) and the corresponding deviations (3) in the adjustable time window for measurement signal values subsequent in time, and in that a processor device is provided for obtaining an evaluation quantity which actuates the alarm device when the evaluation quantity exceeds an adjustable alarm parameter.

Images