DE102007012783B4 - Method for analyzing a measuring signal, in particular of a medical image detector, for extraordinary deviations - Google Patents

Abstract

Method for analyzing a measurement signal (S) consisting of individual signal values (S (k)) for extraordinary deviations, in particular of a measurement signal (S) from a medical image detector (8), the measurement signal (S) with a detector having multiple channels (k) (8) has been recorded and each channel (k) is assigned a respective signal value (S (k)), with the measurement signal (S) being preprocessed in a first step and the preprocessed signal (SK) for determination in a second step of a filter signal (TP) is subjected to filtering, in a third step the measurement signal (S) and the filter signal (TP) are compared with one another, with a signal group (Sg (m)) in the measurement signal (S) being determined by the preprocessing has an extraordinary deviation and the measurement signal (S) is corrected in the area of the signal group (Sg (m)) in such a way that an artifact (AF) in the filter signal (TP (SK)), especially in the edge area of adjacent signalg ruppen is avoided.

Description

The invention relates to a method for analyzing a measurement signal for extraordinary deviations.

In the following, reference is made to a measurement signal which has been detected by a detector comprising several channels. This is in particular a medical image detector, as used for example in computed tomography for the acquisition of medical image information. Each channel of the detector is assigned a signal value.

By means of an analysis of the measurement signal, signal tendencies and thus extraordinary deviations of the measurement signal can be detected. These extraordinary deviations make it possible to draw conclusions about the technical condition of the detector. In particular, when performing a service on the detector, faulty channels can be detected and subsequently measures taken to repair the detector. Another application is to detect deviations in the signal values of the individual channels with one another based on manufacturing tolerances or on an aging process.

To determine the extraordinary deviations, the detector is subjected to a standardized event. An image detector associated with a computer tomograph is an exposure to X-radiation of a given intensity. In this case, a quasi-continuous measurement signal is generated, d. H. Normally, the signal values of adjacent channels differ only slightly from each other.

The starting point for determining the signal tendency of a measuring signal is in particular a so-called low-pass filtering of the signal values. In low-pass filtering, the signal values are usually subjected to local, polynomial, rectangular or Gaussian averaging. In contrast to a pure moving averaging, a weighting of the individual signal values additionally takes place via a so-called filter core which, for example, has the form of a discrete Gaussian distribution function in Gaussian averaging.

Subsequently, a comparison of the filter signal with the measurement signal is made. This will be explained by way of example using the example of low-pass filtering. In this case, the comparison of the low-pass filtered signal with the measurement signal is performed by subtraction. In other words, the comparative component is subtracted from the signal values of the measurement signal. This difference is called high-pass filtering, and its result is a high-pass filtered signal. The high-pass filtered values deviate noticeably from zero only in areas of extraordinary deviation.

Since the values of the low-pass filtering are calculated by an averaging, an area with an extraordinary deviation of its signal values always also acts on adjacent channels. It comes with the high-pass values to so-called artifacts in the adjacent channels, which simulate there also an extraordinary or at least significant deviation of the signal values. Therefore, an extraordinary deviation in this way is not exact, but can be localized only with a certain blurring.

From the DE 196 81 690 T5 (= DE 196 81 690 T1 A strip rejection filter for computed tomography can be seen in which low and high frequency signals are generated from acquired data which are subjected to filtering, the low frequency signals being combined with the filtered signals to produce a strip correction signal.

From the DE 101 55 089 C1 is a method for removing rings and partial rings in computed tomography images, in which a correction sinogram is determined for the suppression of artifacts in computer tomography raw data, which is subtracted from a measured output sinogram. For this purpose, the measured output sinogram is subjected to both high-pass filtering and low-pass filtering, and the amounts of the gradients of the individual data points of the sinogram are determined, with data points exceeding a threshold value being removed after certain selection criteria.

From the DE 102 31 024 B3 is a method for exchanging detector modules in an X-ray detector of a computer tomograph refer to correction tables are adapted to eliminate temperature-dependent signal changes.

The invention has for its object to improve the detection of signal tendencies of a measurement signal and thus the localization of extraordinary deviations of the measurement signal.

For this purpose, the measurement signal is subjected to preprocessing in a first step. Only in a second step is the preprocessed signal for determining a filter signal subjected to suitable filtering. In a third step, the measurement signal and the filter signal are compared with each other. This is done, for example, by the formation of the difference between the measurement signal and the filter signal. Thus, in a high-pass filtering, as already described, the low-pass filtered signal is subtracted from the measurement signal.

The pre-processing determines a signal group in the measurement signal which has an extraordinary deviation. A signal group is henceforth to be understood as meaning a plurality of channels of the detector with their signal values, which computationally form a unit for the analysis. The measurement signal is corrected in the region of the signal group in such a way for determining the preprocessed signal that an artifact of the filter signal in the edge region of the signal group is avoided. It is crucial that pre-processing takes place before the filtering and that the filter-specific artifacts are anticipated by the preprocessing by a suitable correction. A preprocessing or correction is always to be understood here as meaning that the signal values of the signal group are replaced by the corrected values. Therefore, in the sequence often shortened by the correction of the measuring signal is spoken. The corrected values are selected in such a way that they correspond to a signal course that can be expected without the extraordinary deviation. In the preprocessed signal the extraordinary deviation is thus compensated.

The occurrence of an artifact, as would occur in the event of an extraordinary deviation in the edge region of the signal group without preprocessing, is thus avoided. The signal group with the extraordinary deviation can therefore be exactly localized. This makes it possible to determine exactly which channels of the detector have the extraordinary deviation. A maintenance measure to eliminate the extraordinary deviation is thus carried out targeted. The elimination of the extraordinary deviation is therefore quick and inexpensive feasible. Thus, a medical diagnostic device with an image detector by a service technician on site, so in a doctor's office or in a department of a hospital, quickly repair. An examination plan therefore comes only slightly in time delay. Unnecessary waiting times for patients to be examined do not arise.

Appropriately, a median filtering is performed to correct for deviations of individual signal values. In median filtering, the so-called median or central value is determined. The median is the number value that is in the order of a number of numerical values in the middle, d. H. one half of the numbers are less than or equal to the median and the other half of the numbers are greater than or equal to the median. The median width denotes the number of numerical values considered. For an even median width or an even number of signal values, the median is the mean of the two middle numbers.

Beispielsweise ist Md[5](2, 7, 3, 6, 5) = 5.This yields for the median Md with the median width j, based on the numerical values z ₁ ,..., Z _j :

For example, Md [5] (2, 7, 3, 6, 5) = 5.

In median filtering, therefore, each signal value is replaced by its median value with a predetermined median width. In this case, a number of adjacent signal values S (k '), k' ≠ k is used to determine the median of a signal value S (k). It makes sense to include 1 signal values S (k ') left and right of the signal value S (k) in the median filtering. The median is thus determined for a sequence of numbers with an odd number of values. The signal value S (k) is replaced by its median, formally: Ŝ (k) = Md [2l + 1] (S (k-1), ..., S (k), ..., S (k + 1)).

The median filtering is performed for each signal value in turn. It is therefore a sliding median filtering. For the first and last signal values, the median is extrapolated. Compared to the arithmetic mean (average), the median has the advantage of being more robust against outliers. Thus, with a median width of 3, an outlier can be compensated, with a media width of 5, two outliers can be compensated. Such an outlier is therefore not considered for further consideration, but replaced by a typical adjacent signal value. If only individual outliers, so-called single-channel outliers, exist with regard to the signal values, they are therefore already eliminated by means of the computationally simple median filtering with only a slight falsification of the measurement signal. The signal values are thus completely retained except for the outliers, so that, in contrast to a moving averaging, after the median filtering almost the measurement signal is present.

Advantageously, the individual signal values are combined into signal groups. A signal group here is to be understood as meaning a plurality of channels of the detector with their signal values, which are regarded as one unit. The signal values contained in the signal groups are evaluated together to determine a signal group with an extraordinary deviation. By combining signal values in a signal group, subsequently, fewer and simpler arithmetic operations are necessary in order to find a signal group with an extraordinary deviation than when the individual signal values are considered. A computational analysis of the signal values can thus be carried out quickly without great effort. It allows limiting the extraordinary deviation to a specific signal group.

Suitably, the individual channels assigned to the signal group form a coherent unit. In other words, the signal group comprises adjacent channels with their signal values. In this way, a particularly simple summary of each channel in turn is feasible. In particular, this contiguous unit is a hardware unit or an assembly of the detector. In this way, the extraordinary deviation can be clearly assigned to the hardware unit or the module.

In an advantageous variant, an average value of the signal values is formed in each case for the individual signal groups. An increment value for one of the two signal groups is determined from the mean values of two adjacent signal groups. In this case, the increment value is understood in particular to mean the difference between the two mean values. It can also be any other mathematical operation.

In particular, a calculation of the increment value as difference of the mean values of adjacent signal groups is easy to carry out. Such an increment value thus contains information about the deviation of the mean values of the signal groups from one another. An incremental value which is particularly high in terms of its amount thus provides an indication that an extraordinary deviation exists in one of the two adjacent signal groups.

In an expedient development, the increment value is compared with an increment value characteristic of the measurement signal and classified as good or critical depending on a decision rule. Classifying an increment as good means that it will not undergo further investigation. The classification of an increment as critical means that one of the two signal groups corresponding to the increment can be considered as a candidate for extraordinary deviation.

Such characteristic increment values can also be calculated from the signal values of the measurement signal in a simple manner. As a decision rule, for example, a simple arithmetic comparison can be implemented by means of comparison operators, in which it is checked whether the increment value is less than, equal to or greater than the characteristic increment value. In this case, the characteristic increment value may also be a functional one.

B:

charakteristischer Inkrementwert

ISM(i):

Inkrementwert der i'ten Signalgruppe

Md[N]:

Median mit der Medianbreite N

mit N gleich der Zahl sämtlicher betrachteter Inkremente diese Aufgabe. Damit ist der charakteristische Inkrementwert ein typischer Inkrementwert für alle Signalgruppen. Der Vergleich des Inkrementwertes mit einem derartigen charakteristischen Inkrementwert liefert demnach eine Aussage, ob der betrachtete Inkrementwert größer ist als der typische globale Inkrementwert aller Signalgruppen, und demnach sind beide korrespondierenden Signalgruppen des betrachteten Inkrements Kandidaten für eine außerordentliche Abweichung.In one variant, the characteristic increment value is a global increment value. For example, the increment value is fulfilled B = Md [N] (ISM (1), ..., ISM (N))

B:

characteristic increment value

ISM (i):

Increment value of the i'th signal group

Md [N]:

Median with the median width N

with N equal to the number of all considered increments this task. Thus, the characteristic increment value is a typical increment value for all signal groups. The comparison of the increment value with such a characteristic increment value thus provides a statement as to whether the considered increment value is greater than the typical global increment value of all signal groups, and thus both corresponding signal groups of the considered increment are candidates for an extraordinary deviation.

A(m):

charakteristischer Inkrementwert

ISM(i):

Inkrementwert der i'ten Signalgruppe

Md[nl]:

Median mit der Medianbreite nl, wobei nl eine ungerade Zahl ist

In another variant, the characteristic increment value is a local increment value. This is fulfilled by, for example A (m) = Md [nl] (ISM (m-nl / 2), ..., ISM (m), ..., ISM (m + nl / 2))

At the):

characteristic increment value

ISM (i):

Increment value of the i'th signal group

Md [nl]:

Median with the median width nl, where nl is an odd number

Thus, the characteristic increment value is a typical increment value in the neighborhood nl of the considered increment value. A comparison of the increment value with the characteristic increment value thus provides a statement as to whether the considered increment value is greater than the typical local increment value in the environment and therefore its two corresponding signal groups represent candidates for an extraordinary deviation.

GI(m):

charakteristischer Inkrementwert

α, β:

Linearkoeffizienten

A(m):

lokaler Inkrementwert

B:

globaler Inkrementwert

α und β sind empirische Konstanten. Dabei werden der lokale und der globale Inkrementwert beispielsweise gebildet, wie bereits weiter oben beschrieben. Auf diese Weise lässt sich der Inkrementwert mit einem charakteristischen Inkrementwert vergleichen, der einen lokalen und einen globalen Anteil aufweist, und der somit eine besonders treffende Information über den Verlauf des Messsignals aufweist. Auch dieser Vergleich dient dem Auffinden von Signalgruppen als Kandidaten für eine außerordentliche Abweichung.In an expedient development, the characteristic increment value is a linear combination of a local increment value and a global increment value. This is fulfilled by, for example GI (m) = | α · A | + β · B

GI (m):

characteristic increment value

α, β:

linear coefficients

At the):

local increment value

B:

global increment value

α and β are empirical constants. In this case, the local and the global increment value are formed, for example, as already described above. In this way, the increment value can be compared with a characteristic increment value which has a local and a global component, and thus has particularly accurate information about the course of the measurement signal. This comparison also serves to find signal groups as candidates for extraordinary deviation.

HA:

charakteristischer Amplitudenwert

S(k):

Signalwert

k:

Kanal

N:

Kanalanzahl

max[S(k)]:

maximaler Signalwert

min[S(k)]:

minimaler Signalwert

ha1:

Offset

ha2:

Skalierungsfaktor, normalerweise ha2 << 1

ha1 und ha2 sind empirische Konstanten. Der charakteristische Amplitudenwert ist ein Maß für die Spreizung der Signalwerte aller Kanäle und über den Skalierungsfaktor ha2 << 1 ein Maß für die typische Spreizung der Signalwerte innerhalb einer Signalgruppe. Durch den Vergleich eines Inkrementwertes mit dem charakteristischen Amplitudenwert lässt sich ermitteln, ob der Inkrementwert unter oder über dieser Spreizung liegt und somit die korrespondierenden Signalgruppen einen Kandidaten für eine außergewöhnliche Abweichung darstellen.In an expedient variant, the increment value is compared with an amplitude value characteristic of the measurement signal and classified as good or critical depending on a decision rule. Such a characteristic amplitude value is in the simplest case the difference between the maximum and the minimum signal value, based on the signal values of all channels. This difference value can be further scaled by scaling according to HA = ha1 + ha2 * (max [S (k)] - min [S (k)]), k = 1, ..., N

HA:

characteristic amplitude value

S (k):

signal value

k:

channel

N:

number of channels

max [S (k)]:

maximum signal value

min [S (k)]:

minimum signal value

ha1:

offset

ha2:

Scaling factor, usually ha2 << 1

ha1 and ha2 are empirical constants. The characteristic amplitude value is a measure of the spread of the signal values of all channels and, via the scaling factor ha2 << 1, a measure of the typical spread of the signal values within a signal group. By comparing an increment value with the characteristic amplitude value, it is possible to determine whether the increment value lies below or above this spread and thus the corresponding signal groups represent a candidate for an extraordinary deviation.

In an advantageous variant, the increment value of the signal group is compared with a scatter value characteristic of the signal group and classified as good or critical depending on a decision rule. For example, this characteristic scatter value is the standard deviation of the signal values within a signal group from its mean value. Here it is checked whether the increment value lies within a distribution band characteristic of the signal group.

The result of all decision rules is expediently stored in a correlation matrix. This is done in particular by the correlation matrix getting an entry for each increment value. In particular, this entry is equal to zero if the increment value is good and nonzero if the increment value is classified as critical. In this way, the evaluations of all increment values can be easily stored in the correlation matrix. If the increment value has additional information, for example a positive or a negative sign resulting from subtraction, then this additional information can be stored in the correlation matrix either by means of different numerical values or by means of a sign constraint.

In an advantageous development, an increment value is classified as defective in the correlation matrix if the increment value was classified as critical in the case of several decision rules. This erroneous classification corresponds to a uniquely identifiable entry in the correlation matrix. By means of different numerical values, it is also possible to store the information for which decision rule the incremental value was classified as faulty. A classification as faulty means that an extraordinary deviation exists in one of the two signal groups corresponding to an increment value according to the classification according to the decision rules.

Advantageously, an increment value is already classified as good in the correlation matrix if it was classified as good when compared with the characteristic increment value. A comparison with a characteristic scattering value or with a characteristic amplitude value does not take place in this case. A comparison of an increment value with the characteristic increment value thus forms the sharpest criterion in the evaluation of a signal group. Already with the execution of a single comparison operation, the largest part of the increment values can be finally evaluated as good, assuming a typical course of the measurement signal, in which no or only a few extraordinary deviations occur. Only a few incremental values classified as critical when compared with the characteristic increment value must be subjected to further consideration. The processing costs are kept within limits.

In an expedient development, a second correlation matrix is created from the first correlation matrix, wherein different types of extraordinary deviations are deduced by comparison of the adjacent values in the first correlation matrix and type-specific indices are assigned to the corresponding signal groups in the second correlation matrix.

In this way, by means of the second correlation matrix it is possible to specify exactly how a correction of a marked signal group is to be carried out.

In an expedient variant, as signal groups which have an extraordinary deviation, only those signal groups whose increment value has been classified at least as critical are identified. Thus, the critical classification is a necessary but not sufficient condition for the existence of an extraordinary deviation. By means of a further analysis of the entries of the first correlation matrix corresponding to an increment value, it is then decided whether there actually exists a criterion for the presence of an extraordinary deviation.

In an expedient development, as signal groups which have an extraordinary deviation, only those signal groups are identified to which a code number is assigned. This code number is entered for an increment value either at the corresponding position in the first correlation matrix or in the second correlation matrix. This makes it possible to understand in a simple and comprehensible manner on the basis of the code number whether or not there is an extraordinary deviation for an incremental value of its corresponding signal group.

Suitably, the or each signal group is corrected with an extraordinary deviation in which a corrected measurement signal for the signal values of the signal group is calculated by means of a calculation method by means of values from surrounding signal groups. The corrected signal values result in particular from the signal values of the signal groups found to be good by extrapolation and / or intrapolation and / or averaging. The signal values of the signal group with the extraordinary deviation are replaced by the corrected signal values. For the signal groups which have no extraordinary deviation, the original signal values are retained. Finally, a piecewise correction is performed only for the signal groups with the extraordinary deviations whose signal values are generated, for example, in the manner of a polygon. In this way, the corrected signal resembles the measurement signal that would have been expected without an extraordinary deviation. Thus leads a Low pass with a subsequent calculation of the high pass from the original measurements and the low pass values of the corrected signal not to artifacts in the periphery of the signal groups affected by extraordinary deviation. The extraordinary deviation can thus be read from the high-pass values and assigned exactly to a signal group. An erroneous evaluation of a group of signals adjacent to this signal group as also provided with an extraordinary deviation is thus reliably avoided.

If the detector is subdivided into modules and a signal group combines the respective channels of a module, advantageously an identified signal group can be assigned to its corresponding module in the detector. The module can also be divided into further subunits, for example in so-called ASICs. A module here is to be understood in each case as an apparatus having an electrical circuit which, for a plurality of channels of the detector, effects a detection of an event and a conversion of this event into a proportional signal. ASIC is the name given to an application-specific integrated circuit, in German for an application-specific integrated circuit. This is an electronic circuit that has been realized as an integrated circuit. ASICs are manufactured by many manufacturers according to the customer's requirements, ie they are integrated circuits specially adapted to a task.

Thus, in particular, an assignment of the extraordinary deviation to a detector module or to a subunit of a detector module is possible. It is possible to pinpoint a possible hardware defect. Optionally, taking further action, such as replacing the complete detector module or one of its subunits, is possible. These measures can be implemented quickly due to the hardware limitation of the extraordinary deviation. They are limited to maintenance or replacement of the corresponding hardware unit. Optionally, a mathematical correction of the extraordinary deviation, as already described, take place.

In another variant, a correction of detector values is carried out by means of the corrected signal. Thus, on the one hand, detectable as extraordinary deviations detectable manufacturing tolerances between individual channels of the detector or between detector modules. On the other hand, correction values can be calculated by means of periodically performed calibration measurements, which correct for uneven aging of the detector.

For a signal group with such an extraordinary deviation, correction values are stored in a table. By means of these correction values, the signal values of the signal group are corrected in a subsequent measurement. Such a check of the detector for extraordinary deviations is simple and can be carried out periodically with little effort. The service life of the detector can thus be significantly extended. If the extraordinary deviations are considered to be no longer correctable, a maintenance action of the detector can then be carried out selectively, as already described. Although the correction values could also be determined without preprocessing the measurement signal from the high-pass values, for the reasons already described, these correction values are subject to a significantly higher error.

Advantageously, the method can be used with an X-ray detector, in particular for a computer tomographer. By means of the method, the X-ray detector can be quickly checked for possible errors. An error which manifests itself in an extraordinary deviation in the signal values of a detector module can be quickly eliminated by exchanging this detector module. Since the extraordinary deviation can be assigned exactly to a detector module, such an exchange can be carried out comparatively quickly. A test of the X-ray detector can be carried out very quickly by means of a defined X-ray radiation acting on the X-ray detector. Overall, the safe localization of a faulty module results in a short maintenance interval.

An embodiment of the invention will be described in more detail with reference to a drawing. The individual figures show:

1 CT Scanner

2 Operating principle of an x-ray detector

3 Distribution of the X-ray detector in detector modules

4 process scheme

5a -B Examples of Creating a Correlation Matrix

6a -E first application example

7a -E second application example

8a -E third application example

According to 1 has a computer tomograph 2 a gantry 4 on, in which an X-ray source 6 and the X-ray source 6 opposite to an x-ray detector 8th are rotatably mounted. In the gantry 4 is by means of a longitudinal direction 10 movable patient bed 12 a patient 14 inserted for examination. The patient 14 is by means of the patient bed 12 longitudinal 10 positioned so that its body region to be examined 16 between X-ray source 6 and x-ray detector 8th to come to rest. By means of a control unit 18 becomes the X-ray source 6 for generating an X-ray 20 driven. The x-ray 20 radiates the body region 16 the person 14 and is determined by means of the X-ray detector 8th detected. By means of a co-rotation of x-ray source 6 and x-ray detector 8th becomes the body region 16 from different positions with the x-ray 18 for generating a measurement signal S through. This measurement signal S is from a processing unit 22 detected. All measurement signals S from a complete co-rotation of the X-ray source 6 and x-ray detector 8th be in the processing unit 22 cached and then converted into image information BI. The image information BI becomes a sectional image of the body region 16 on one as a monitor 24 trained display element displayed. For the acquisition of a further image information BI and the generation of a further sectional image is the patient bed 12 longitudinal 10 moved, and it is carried out a new measurement.

In 2 is the operating principle of the X-ray detector 8th shown schematically. The x-ray radiation 20 meets after seeing the body region 16 the person to be examined 14 has irradiated on a trained as a surface element scintillator 26 , There is the X-ray radiation 18 in light signals 28 transformed. At the x-ray 20 opposite side of the scintillator 26 is a variety of photodiodes 30 arranged. Each of these photodiodes 30 corresponds to a channel k of the X-ray detector 8th , Overall, the X-ray detector has 8th a number of N photodiodes 30 and thus a number N of channels k. Each photodiode 30 converts the light signal that acts on it 28 into a proportional signal value (S (k)). The measurement signal S of the X-ray detector 8th is the totality of all these N signal values S (k).

3 shows a section of the detector 8th from his the X-ray 20 facing side. The detector 8th is in detector modules 32 divided, which is the detector 8th longitudinal 34 and in the transverse direction 36 divide into individual units. Each detector module in turn is in two ASIC modules 38 divided. An ASIC module 38 are eight photodiodes 30 and thus allocated channels k. Thus, a detector module comprises 32 sixteen photodiodes 30 and thus channels k.

In the 3 are twelve detector modules 32 with two ASIC modules each 38 shown. This corresponds to a total of 192 channels k.

For a test of the X-ray detector 8th on its functioning, the X-ray detector 8th from the control unit 18 subjected to a predetermined X-radiation. There is no person 14 between the X-ray source 6 and the X-ray detector 8th , so that the exposure of the X-ray detector 8th with the x-ray 20 with a uniform power density occurs.

The x-ray detector 8th generates the measurement signal S from the X-ray radiation. The measurement signal S has a signal value S (k) for each of the N channels k and is an N-tuple. This measurement signal S is from the processing unit 18 according to 4 in a process step 40 detected. Such a measurement signal S is in the 6a . 7a and 8a shown.

In one process step 42 a median filtering Md of the signal values S (k) is performed. For the median Md with the median width j, based on the signal values S ₁ , ..., S _j , the following applies:

The individual channels k with their signal values S (k) are in one process step 44 into so-called signal groups Sg (m). In each case, a number of Sb channels k forms a signal group Sg (m). Is formal Sg (m) = {S ((m-1) * Sb + 1), ..., S (m * Sb)}, m ∈ 1, ..., N / Sb.

For further consideration, Sb = 16 or Sb = 8 is set and the channels k are combined such that each signal group Sg (m) is the channels k of a detector module 32 or an ASIC 38 includes.

In one process step 46 the mean value SM (m) of the signal values S (k) contained in the signal group Sg (m) is calculated for each signal group Sg (m):

In one process step 48 is calculated between each two average values SM (m), SM (m + 1) of signal groups Sg (m), Sg (m + 1) adjacent to their index whose so-called increment ISM (m). The increment is defined as the difference between the two mean values SM (m), SM (m + 1): ISM (m) = SM (m + 1) - SM (m); m = 1, ..., N / Sb - 1.

In one process step 50 First, a global increment value B is calculated from all increment values ISM (m) B = Md [N / Sb-1] (ISM (1), ..., ISM (N / Sb-1)) calculated. Furthermore, in one process step 52 according to HA = ha1 + ha2 * (max [S (k)] - min [S (k)]); k = 1, ..., N a characteristic amplitude value HA is calculated. ha1 is an offset, ha2 is a scaling factor with ha2 << 1. ha1 and ha2 are empirically determined coefficients.

Subsequently, the increments ISM (m), m = 1,..., N / Sb-1 in one process step 54 in turn examined whether they represent a candidate for an extraordinary deviation. This is achieved by means of a loop structure.

First, in one process step 54a for the increment ISM (m), a local increment value A (m) in the neighborhood nl of m according to A (m) = Md [nl] (ISM (m-nl / 2), ..., ISM (m), ..., ISM (m + nl / 2)) calculated.

In one process step 54b is calculated from the local increment value A (m) and in the process step 50 calculated global increment value B is a characteristic increment value GI (m) as a linear combination according to GI (m) = | α · A (m) + β · B | calculated. α and β are empirically determined linear coefficients.

In one process step 54c In addition, a characteristic scattering value HS (m) with respect to the signal values S (k) contained in the signal group Sg (m) is formed.

Each signal group Sg (m) has Sb = 16 signal values S (k): Sg (m) = (S1, ..., S16).

There are two mean values m1 = 1/3 · (S2 + S3 + S4); m2 = 1/3 · (S13 + S14 + S15) calculated. Next is HS (m) = hs1 + hs2 · | m1 - m2 / m1 + m2 |.

The characteristic scattering value HS (m) is a measure of a typical fluctuation of the signal values S (k) within the signal group Sg (m). It provides further information than the standard deviation σ (Sg (m)) of the signal values S (k) of a signal group Sg (m), since it also takes into account the position of the signal values S (k) within the signal group Sg (m).

But it can also be a scattering HS '(m) with HS '(m) = hs1' + hs2 '· σ (Sg (m)) taking into account the standard deviation σ (Sg (m)). hs1 'and hs2' are scaling factors by means of which the standard deviation σ (Sg (m)) is adjusted.

Finally, the increment ISM (m) is evaluated on the basis of decision rules ER1, ER2, ER3. According to the first decision rule ER1, it is checked whether the absolute value of the increment ISM (m) is larger than GI (m), d. H. if | (ISM (m) |> GI (m) holds) Furthermore, according to the second decision rule ER2, it is checked whether the absolute value of the increment ISM (m) is greater than HA, ie, | ISM (m) |> HA. Finally, according to the third decision rule ER3, it is checked whether the absolute value of the increment ISM (m) is greater than HS (m), ie, | ISM (m) |> HS (m).

In another variant, the third decision rule ER3 is modified. In this case, the dispersion HS (m) of the values within a signal group Sg (m) is compared with the value HL, i. H. it is checked if HS (m)> HL (m).

Where HL (m) is too close HL (m) = HL '(m) * f (HA) and HL '(m) too HL '(m) = hl * Md [2 * nm + m] (HS (m - nm), ..., HS (m), ..., HS (m + nm)) defined with nm = 1, 2, .... hl is an empirical constant. HL (m) is a typical scattering value HL '(m) for the environment of the signal group Sg (m), weighted with a functional f (HA) dependent on the characteristic amplitude value HA.

In practice, for example, the choice f (HA) = HA + 4 / HA + 50 , hl = 10 and with nm = 1 HL '(m) = 10 * Md [3] (HS (m-1), HS (m), HS (m + 1)) or with nm = 2 HL '(m) = 10 * Md [5] (HS (m-2), HS (m-1), HS (m), HS (m + 1), HS (m + 2)) satisfactory results. In the 4 is the alternative decision rule ER3 dashed lines.

Only if the considered increment ISM (m) is classified as critical with respect to all three decision rules ER1, ER2, ER3, it is classified as critical as a whole and thus is a candidate for an extraordinary deviation. This candidate will undergo further analysis. On the other hand, if the increment ISM (m) is not classified as critical with regard to one of the three decision rules ER1, ER2, ER3, it is classified as good overall. Immediately proceed with the evaluation of the following increment ISM (m + 1).

According to process step 60 a correlation matrix Corr1 is provided which receives an entry for each increment ISM (m). The correlation matrix is initially preallocated with Corr1 = (0, ..., 0) as (N / Sb-1) tuple. If all three decision rules ER1, ER2, ER3 apply to a considered increment ISM (m), this increment ISM (m) is classified as critical. As a consequence, the mth entry of the correlation matrix Corr1 becomes too Corr1 (m) = sgn [ISM (m)] set. In this case, sgn [ISM (m)] denotes the sign of the increment ISM (m) resulting from the subtraction of SM (m + 1) and SM (m).

According to process step 54 Then, the following increment ISM (m + 1) is evaluated until all increments ISM (m), m = 1, ..., N / Sb-1 are evaluated.

In accordance with a method step 58 an analysis of the correlation matrix Corr1 is performed. In this case, the value and the sign, in particular of the adjacent coefficients Corr1 (m '), m ≠ m', are checked for each non-zero coefficient Corr1 (m) and therefore for each increment ISM (m) classified as critical. The result is stored in a second coefficient matrix Corr2 as coefficient Corr2 (m). All is valid

The correlation matrix Corr2 contains N / Sb entries, one for each signal group Sg (m). The entry Corr2 (m) describes whether and, if appropriate, how the signal values S (k) of the signal group Sg (m) are to be corrected.

f(Corr1(m), Corr1(m + 1)) bezeichnet dabei ein Funktional, das von den Koeffizienten Corr1(m) und Corr1(m + 1) der Korrelationsmatrix Corr1 abhängt. Es handelt sich hierbei um einen numerischen Wert, der davon abhängt, ob auch Corr1(m + 1) von 0 verschieden ist. Weiterhin wird das Vorzeichen von Corr1(m) und gegebenenfalls von Corr1(m + 1) betrachtet. Der Wert Corr2⁽¹⁾(1) wird durch Extrapolation ermittelt.In the simplest case, only the values Corr1 (m) and Corr1 (m + 1) are checked for Corr1 (m) ≠ 0. This is

f (Corr1 (m), Corr1 (m + 1)) denotes a functional which depends on the coefficients Corr1 (m) and Corr1 (m + 1) of the correlation matrix Corr1. It is a numerical value that depends on whether Corr1 (m + 1) is different from 0. Furthermore, the sign of Corr1 (m) and optionally of Corr1 (m + 1) is considered. The value Corr2 ⁽¹⁾ (1) is determined by extrapolation.

∀m = 1, ..., N/Sb – 2. f(Corr1(m – 1), Corr1(m), Corr1(m + 1), Corr1(m + 2)) liefert wieder einen numerischen Wert, der davon abhängt, welche der Koeffizienten von 0 verschieden sind und welches Vorzeichen sie haben. Die Werte Corr2⁽²⁾(1), Corr2⁽²⁾(2) und Corr2⁽²⁾(N/Sb) werden durch Extrapolation ermittelt.In a refinement, the next coefficient values Corr1 (m-1) and Corr2 (m + 2) are also taken into consideration. Formal:

∀m = 1, ..., N / Sb - 2. f (Corr1 (m - 1), Corr1 (m), Corr1 (m + 1), Corr1 (m + 2)) returns a numerical value that returns depends on which of the coefficients are different from 0 and which sign they have. The values Corr2 ⁽²⁾ (1), Corr2 ⁽²⁾ (2) and Corr2 ⁽²⁾ (N / Sb) are determined by extrapolation.

In a further process step 60 the correction S ^{K of} those signal groups Sg (m) takes place, for whose coefficient Corr2 (m) the correlation matrix Corr2 has a value different from 0. For a signal group Sg ^K (m) to be corrected, the following applies formally:

The correction f (Corr2 (m), S (k ')) depends on the one hand on the numerical value of Corr2 (m). It is determined by means of the signal values S (k ') of the adjacent signal groups Sg (m'), m '≠m where Corr2 (m') = 0, ie where there is no extraordinary deviation. The correction can be a combination of averaging, intrapolation and extrapolation.

This allows the corrected signal group Sg ^K (m) to be summarized: Sg ^K (m) = {S ^K ((m-1) * Sb + 1), ..., S ^K (m * Sb)}, m ∈ 1, ..., N / Sb.

For the corrected signal S ^K, this yields: S ↦ S ^K Sg (m) → Sg ^K (m) ∀m: Corr2 (m) ≠ 0 Sg (m) → Sg (m) otherwise

The corrected measurement signal S ^K is thus piecewise composed of areas in which a correction is performed, and of areas in which the measurement signal S is taken over unchanged.

In a further process step 62 a low-pass filtering TP of the corrected measurement signal S ^{K is} performed. In this case, the signal values S ^K (k) are weighted with the filter kernel G (i) of a Gaussian normal distribution. For the filter kernel G (i) of the Gaussian normal distribution, the following applies:

For weighting, 2l + 1 values (filter width) are used. NSigma determines the discrete range of normal distribution used for weighting. In practice z. Eg NSigma = 3 or NSigma = 5 application.

Changing 1 adjusts the character of the weighting. The low-pass filtering TP for a corrected signal value S ^K (k) is given by

The result is the low-pass filtered corrected signal TP (S ^K ).

In a final process step 64 the high-pass filtered values HP (S ^K (k)) from the low-pass filtered values TP (S ^K (k)) is calculated. For this purpose, the low-pass filtered values TP (S ^K (k)) are subtracted from the measured values S (k), HP (S ^K (k)) = S (k) -TP (S ^K (k)); k = 1, ..., N.

The result is the high-pass filtered signal HP (S (k), TP (S ^K )).

5a -B exemplifies the transition from process step 56 to process step 58 and the detection of Corr2 from Corr1. In 5a lies with respect to the signal group Sg (m ^ ₁ ) an extraordinary deviation abw ago. This is from the two increments ISM (m ^ ₁ - 1) = SM (m ^ ₁ ) - SM (m ^ ₁ - 1) and ISM (m ^ ₁ ) = SM (m ^ ₁ + 1) - SM (m ^ ₁ ) seen. With a strong positive deviation Abw the values of the signal group Sg (m ^ ₁ ) is ISM (m ^ ₁ - 1) big and positive with Corr1 (m ^ ₁ - 1) = +1 and ISM (m ^ ₁ ) big and negative with Corr1 (m ^ ₁ ) = -1 , Due to the two adjacent entries in Corr1 and the change of sign, the extraordinary deviation Abw is thus clearly identified. It will Corr2 (m ^ ₁ ) = f (Corr1 (m ^ ₁ - 1), Corr1 (m)) = k ₁ set. Thus, the type of the signal group is Sg (m ^ ₁ ) to be performed correction k ₁ predetermined.

A negative extraordinary deviation can be detected in analogous form.

5b shows an extraordinary deviation abw, referring to two adjacent signal groups Sg (m ^ ₂ ) and Sg (m ^ ₂ + 1) extends. Here is ISM (m ^ ₂ - 1) = SM (m ^ ₂ ) - SM (m ^ ₂ - 1) large, positive → Corr1 (m ^ ₂ ) = +1 ISM (m ^ ₂ ) = SM (m ^ ₂ + 1) - SM (m ^ ₂ ) in the normal range → Corr1 (m ^ ₂ + 1) = 0 ISM (m ^ ₂ + 1) = SM (m ^ ₂ + 2) - SM (m ^ ₂ + 1) large, positive → Corr1 (m ^ ₂ + 2) = -1.

Are also the adjacent entries in Corr1, namely Corr1 (m ^ ₂ - 1) and Corr1 (m ^ ₂ + 3), equal to 0, so can from the condition Corr1 (m ^ ₂ ) + Corr1 (m ^ ₂ + 2) = 0 be closed to an extraordinary deviation Abw which the two signal groups Sg (m ^ ₂ ) and Sg (m ^ ₂ + 1) concerns. Thus, will Corr2 (m ^ ₂ ) = k ₂ and Corr2 (m ^ ₂ + 1) = k ' ₂ . The coefficients k ₂ and k ' ₂ specify the type of correction.

Analog can be from the condition Corr1 (m ^ ₂ ) + Corr1 (m ^ ₂ + 2) = 0 such as Corr1 (m ^ ₂ ) = -1 and Corr1 (m ^ ₂ + 2) = +1 on one the two signal groups Sg (m ^ ₂ ) and Sg (m ^ ₂ + 1) exceptional negative deviation.

6a shows a measurement signal S of the X-ray detector 8th , The measurement signal S has an extraordinary deviation Abw of the signal values S (k) in the signal group Sg (m ₁ ). With regard to the other signal groups Sg (m), m ≠ m ₁ , the measuring signal S runs uniformly with only slight fluctuations. The measurement signal S is in accordance with the in the 4 analyzed algorithm. In this case, the extraordinary deviation Abw of the signal group Sg (m ₁ ) is detected. The signal values S (k) of the signal group Sg (m ₁ ) are replaced by corrected signal values S ^K (k). In this case, the signal values S (k) of the adjacent signal groups, in particular of Sg (m ₁ - 1) and Sg (m ₁ + 1), are used for the correction, from which the corrected signal values S ^K by means of a combination of interpolation, extrapolation and averaging (k) are generated for Sg (m ₁ ). This results in a polygon, which in the 6b is shown as a dashed line. The correction is performed such that the corrected signal values S ^K (k) correspond to the signal values to be expected for the signal group Sg (m ₁ ) without the extraordinary deviation Abw. Outside the signal group Sg (m ₁ ), the signal values S (k) are not corrected. Thus, the corrected signal S ^K corresponds to the original measurement signal S except for the region of the signal group Sg (m ₁ ).

6c shows the low-pass filtered corrected signal TP (S ^K ). Due to the correction before the low-pass filtering TP, a low-pass signal TP (S ^K ) which is uniform even in the region of the signal group Sg (m ₁ ) results. Without the correction of the measurement signal S, the extraordinary deviation Abw in the low-pass signal TP (S) would still be clearly recognizable, although attenuated by the low-pass filtering TP.

6d shows the high-pass filtering HP from the low-pass signal TP (S ^K ) and the measured signal S calculated high-pass signal HP (S ^K ). Outside of Sg (m ₁ ), there is only one background due to the subtraction. The deviation Abw can thus be clearly determined and assigned to the signal group Sg (m ₁ ).

Relative to the x-ray detector 8th of the computer tomograph 2 this means the following: each signal group Sg (m) corresponds to the channels k of a detector module 32 , The detected extraordinary Deviation Abw for the signal group Sg (m ₁ ) corresponds to a defect in the detector module 32 with the number m ₁ . As a result, targeted measures for the repair of the corresponding detector module 32 to be hit. This can be, for example, an exchange of this detector module m ₁ .

6e shows for comparison the high-pass signal HP (S) without performing a signal correction before the low-pass filtering. In other words, here only an averaging with a Gaussian filter core takes place in order to generate the low-pass signal, as in method step 66 described. The deviation Abw also clearly shows here. However, the signal values S (k) of the adjacent signal groups Sg (m ₁ -1), Sg (m ₁ + 1) also have a large deviation. This strong deviation is an artifact Af and comes about only through the way of averaging. As a result, however, the detector modules m ₁ - 1, m ₁ and m ₁ + 1 are identified as defective. On the one hand, therefore, the detector modules m ₁ - 1 and m ₁ + 1 are examined as part of a maintenance for alleged errors out. In extreme cases, all three detector modules m ₁ - 1, m ₁ and m ₁ + 1 exchanged. Maintenance of the X-ray detector 8th Therefore takes longer than in a preprocessing of the measurement signal S for detecting signal tendencies. In addition, in extreme cases, the detector modules m ₁ - 1 and m ₁ + 1 exchanged, although they have no error. This leads to increased material costs during maintenance.

bleiben wiederum unverändert. Das korrigierte Signal S^K ist in 7b dargestellt. Es verläuft nach der Korrektur quasikontinuierlich. Daher lässt sich das korrigierte Signal S^K einer Tiefpassfilterung TP gemäß 7c unterziehen, ohne dass im tiefpassgefilterten Signal TP(S^K) Änderungen erkennbar wären.In 7a is a second measurement signal S of the X-ray detector 8th shown. The measurement signal S has an extraordinary deviation Abw with respect to its signal values in three adjacent signal groups Sg (m ₂ -1), Sg (m ₂ ) and Sg (m ₂ + 1). This extraordinary deviation Abw is detected. Including in particular the adjacent signal groups Sg (m ₂ - 2) and Sg (m ₂ + 2), a calculation of corrected signal values S ^K (k) for the three signal groups Sg (m ₂ - 1), Sg (m ₂ ) and Sg (m ₂ + 1), by means of which the original signal values S (k) are replaced. All other signal groups

remain unchanged. The corrected signal S ^K is in 7b shown. It proceeds quasi-continuously after the correction. Therefore, the corrected signal S ^{K of} a low-pass filtering TP according to 7c without any changes being detectable in the low-pass filtered signal TP (S ^K ).

The three signal groups Sg (m ₂ -1), Sg (m ₂ ) and Sg (m ₂ + 1) with the extraordinary deviations are again in the high-pass filtered signal HP (S ^K ) according to 7d recognizable. From this an error in the three adjacent detector modules m ₂ - 1, m ₂ and m ₂ + 1 can be concluded and corresponding countermeasures can be taken.

If, on the other hand, a low-pass filtering TP is carried out without preprocessing the measuring signal, as in FIG 7e are shown, so the adjacent signal groups Sg (m ₂ - 2) and Sg (m ₂ + 2) are provided provided with an exceptional deviation Abw.

8a shows a third measurement signal S of the X-ray detector 8th with two deviations The first deviation Abw extends over the entire first signal group Sg (1). The second deviation Abw is a deviation limited to a single channel within the signal group Sg (m ₃ ). 8b shows the corrected signal S ^K. The correction Sg ^K (1) for the first signal group is obtained by extrapolating by means of the signal values S ₃ (2) of the second signal group. The second deviation Abw in the signal group Sg (m ₃ ) is already using the median value filtering by method step 42 eliminated. Again, a low-pass filtering TP is performed. The low-pass filtered signal TP (S ^K ) is in 8c shown. The two extraordinary deviations Abw are no longer noticeable in the low-pass filtered signal TP (S ^K ). The high-pass filtered signal HP (S ^K ) is in 7d shown. Clearly, the two deviations Abw in the signal groups Sg (1) and Sg (m ₃ ) can be seen. The preprocessing of the measurement signal S thus also helps in detecting an extraordinary deviation Abw at the edge of the channel spectrum and in detecting the deviation Abw of individual channels.

In 8th In contrast, the high-pass-filtered signal HP (S) is shown without preprocessing the measurement signal S before the low-pass filtering TP. Although the first deviation Abw in the signal group Sg (1) is detected, it is also related to the second signal group Sg (2). The second deviation Abw in the signal group Sg (m ₃ ) is distributed over several channels k due to the low-pass filtering TP. If this single-channel deviation lies at the edge of the signal group Sg (m ₃ ), or if it has a very high value, the adjacent signal groups Sg (m ₃ -1) and / or Sg (m ₃ + 1) are also identified as defective. In particular, single-channel deviations in an X-ray detector 8th be easily corrected by an image processing software and in this case an exchange of a detector module 32 is not absolutely necessary, up to three detectors m ₃ - 1, m ₃ and m ₃ + 1 are unnecessarily exchanged here in the extreme case when performing a maintenance.

As already in the description of 4 addressed, by the choice of the signal width to Sb = 8 also an ASIC 38 assigned signal group Sg (m) and thus the state of a single ASIC 38 Check for abnormal deviation with the procedure described.

Claims (21)

Method for analyzing a measuring signal (S) consisting of individual signal values (S (k)) for extraordinary deviations, in particular of a measuring signal (S) of a medical image detector ( 8th ), wherein the measuring signal (S) with a multiple channels (k) having detector ( 8th ) and a respective signal value (S (k)) is assigned to each channel (k), wherein in a first step the measurement signal (S) is subjected to preprocessing, in a second step the preprocessed signal (S ^K ) for determining a Filtering (TP) is subjected to a filtering, in a third step, the measurement signal (S) and the filter signal (TP) are compared with each other, wherein the preprocessing a signal group (Sg (m)) in the measurement signal (S) is determined, the one has extraordinary deviation and in the region of the signal group (Sg (m)), the measurement signal (S) is corrected, such that an artifact (AF) in the filter signal (TP (S ^K )), in particular in the edge region of adjacent signal groups is avoided. Method according to Claim 1, wherein a median filtering (Md [j]) is carried out for the correction of deviations (Abw) of individual signal values (S (k)). The method of claim 1 or 2, wherein the individual signal values (S (k)) to signal groups (Sg (m)) are combined and evaluated together to determine the signal group (Sg (m)) with the extraordinary deviation Abw. Method according to Claim 3, in which the individual channels (k) assigned to the signal group (Sg (m)) form a coherent unit. Method according to Claim 3 or 4, wherein an average value (SM (m)) of the signal values (S (k)) is formed in each case relative to the individual signal groups (Sg (m)) and from the mean values (SM (m)) of two adjacent signal groups (Sg (m), Sg (m + 1)) an increment value (ISM (m)) for one of the two signal groups (Sg (m)) is determined. Method according to Claim 5, wherein the increment value (ISM (m)) is compared with an increment value (B, A (m), GI (m)) characteristic of the measurement signal and classified as good or critical in dependence on a first decision rule (ER1) becomes. The method of claim 6, wherein the characteristic increment value is a global increment value (B). The method of claim 6, wherein the characteristic increment value is a local increment value (A (m)). The method of claims 6 to 8, wherein the characteristic increment value is a linear combination (GI (m)) of the local increment value (A (m)) and the global increment value (B). Method according to one of Claims 6 to 9, wherein the increment value (ISM (m)) is compared with an amplitude value (HA) characteristic of the measurement signal (S) and is classified as good or critical as a function of a second decision rule (ER2). Method according to one of Claims 6 to 10, wherein the increment value (ISM (m)) of the signal group (Sg (m)) is compared with a scatter value (HS (m)) which is characteristic of the signal group (Sg (m)) and in dependence a third decision rule (ER3) is classified as good or critical. Method according to one of claims 6 to 11, wherein the result of the decision rules (ER1, ER2, ER3) is stored in a correlation matrix (Corr1). Method according to claim 12, wherein in the correlation matrix (Corr1) an increment value (ISM (m)) is classified as erroneous if the increment value (ISM (m)) has been classified as critical in a plurality of decision rules (ER1, ER2, ER3). Method according to claim 12 or 13, wherein in the correlation matrix (Corr1) an increment value (ISM (m)) is classified as good already when compared to the characteristic increment value (A (m), B, GI (m)) was rated as good. Method according to one of Claims 12 to 14, in which a second correlation matrix (Corr2) is created from the first correlation matrix (Corr1), different types of extraordinary deviations being inferred by comparison of the adjacent values in the first correlation matrix (Corr1) the second correlation matrix (Corr2) are assigned to the corresponding signal groups Sg (m) type-specific code numbers (k ₁ , k ₂ , k ' ₂ ). Method according to one of Claims 6 to 15, in which, as signal groups (Sg (m)) which have an extraordinary deviation Abw, only those signal groups (Sg (m)) are identified, their increment value (ISM (m)) being at least critical was classified. Method according to Claims 15 and 16, in which as signal groups (Sg (m)) which have an extraordinary deviation, only those signal groups (Sg (m)) to which a code number (k ₁ , k ₂ , k ' ₂ ) is identified are identified. assigned. Method according to one of Claims 1 to 17, in which the or each signal group (Sg (m)) is corrected with an extraordinary deviation in which by means of signal values (S (k)) from surrounding signal groups (Sg (n), n ≠ m ) is calculated by means of a calculation method, a corrected signal (S ^K ) for the signal values (S (k)) of the signal group (Sg (m)). Method according to one of the preceding claims, wherein a module corresponding to a determined signal group (Sg (m)) ( 32 . 38 ) in the detector ( 8th ), in particular exchanged. Method according to one of the preceding claims, in which a correction of detector values S is carried out by means of the corrected signal (S ^K ). Method according to one of the preceding claims, in which the image detector ( 8th ) to an X-ray detector, in particular for a computer tomograph ( 2 ).

Images