JP4856993B2 - Self-diagnosis type automatic analyzer - Google Patents

Description

  The present invention relates to an automatic analyzer that performs qualitative and quantitative analysis of biological samples such as blood and urine, and particularly in an automatic analyzer that has a function of measuring changes in physical properties of a sample in time series, Further, the present invention relates to an automatic analyzer having a function of checking an operation mechanism or a system that realizes the reaction process.

  For example, in a biochemical analyzer, an automatic analyzer for a clinical test uses a reagent that changes color by reacting with a specific component in a sample, and quantitatively measures the change in color based on a change in absorbance, thereby qualitatively determining the specific component.・ Those that perform quantitative analysis are common. Some of such automatic analyzers have a function of storing a change in absorbance during a reaction process or plotting it on a screen. When an abnormality is found in the measurement result, the operator of the instrument can check whether the sample is really abnormal or whether the abnormal result is occasionally caused by the abnormality of the instrument by examining the plot of the absorbance change. There is sex. In the process of clinical automatic analysis and inspection, abnormalities in the nozzle due to prozone check abnormalities, sampling abnormalities, reagent dispensing abnormalities, stirring mechanism abnormalities, dripping or scattering of high-viscosity reagents, and reagent combinations Although there is a risk that analysis and inspection may not be performed normally due to contamination, crystal precipitation, or the like, these abnormalities may be detected by analysis of reaction process data.

  However, since changes in reaction process data (not limited to absorbance change data) vary depending on analysis items and sample identification, it is difficult to realize a function that automatically determines whether or not there was an abnormality in the reaction process. No one realized it. For this reason, in order to investigate reaction process abnormalities from reaction process data, it is necessary to store a large amount of data, and when an abnormal cause is considered in the analyzer, the reaction process data and the analysis sequence of the apparatus It was necessary to specify the cause of each item in detail, and further, it was necessary to examine individual reaction process data one by one manually, which took time and cost.

Japanese Patent No. 328087 Journal of Quality Engineering Vol.3 No.1: "Comprehensive evaluation and SN ratio by multidimensional information" Technology development in MT system: Japanese Standards Association, Quality Engineering Application Course

  The measurement result in the clinical automatic analysis test only evaluates the final result, so if the measured value is within the set range, the analyzer does not issue an alarm and detection is not possible even if there is an abnormal reaction Met. Moreover, even if the measured value is abnormal, there is no means for quickly determining whether the sample is actually abnormal or whether the sample is due to deterioration or abnormal operation of a mechanism part inside the apparatus.

  The reason why current automatic analyzers cannot quickly determine the presence or absence of abnormalities in the reaction process is largely due to the variety of abnormalities in the reaction process, as described above.

  On the other hand, in recent years, as a method of multivariate data analysis, a method (MT method; Mahalanobis-Taguchi method) that forms a predetermined space at Mahalanobis distance and comprehensively determines whether the data is abnormal is used. Has been. For example, in Non-Patent Document 1, if a Mahalanobis space (reference space) serving as a reference is created based on healthy person data and the Mahalanobis distance calculated for a subject whose health is unknown is smaller than a certain threshold value (for example, 4). This is a method of belonging to a group of healthy people and discriminating as “healthy”, otherwise judging as “unhealthy” or “abnormal”. Such a method has a wide range of applications and has been published in various fields. For example, there are various examples in Patent Document 1 and Non-Patent Document 2.

  If this method can be applied to the determination of the presence / absence of an abnormality in the reaction process, it is considered possible to reduce the possibility that data that is not abnormal is determined to be abnormal or that an abnormal sample is determined to be normal. However, time-series data such as reaction process data is a data group in which the correlation between each data is strong (a constant value within the measurement error and the slope is the same), so the data at each time point is used as a parameter for Mahalanobis. The inverse matrix of the correlation matrix cannot be calculated even when creating a space. For this reason, further data processing such as deleting one of the strongly correlated data (or between measurement items) is required, and it takes time to reconstruct the Mahalanobis space, which is the above-mentioned standard, and the processing efficiency is reduced. There is a potential problem.

  The present invention makes it possible to apply the MT method to data analysis by preliminarily processing reaction process data, so that abnormalities caused by abnormal operation or deterioration in the apparatus during the reaction process can be detected. In addition, it is possible to determine the presence or absence of abnormalities in the reaction process (including abnormalities in the operation of the device, etc.), and to improve the reliability of measurement data and the reliability of the mechanism or subsystem associated with the reaction process An object is to provide an apparatus.

  In order to achieve the above object, the present invention measures the value of a photometer in the reaction process and verifies the analytical measurement result from the reaction process data in order to realize verification of the test result. Furthermore, the analysis results are verified in consideration of abnormalities due to abnormalities and deterioration of the photometer and the mechanism and system associated with each reaction process.

The aforementioned reaction process is a process corresponding to the process sequence of the apparatus up to the final analysis result of the analyzer, and the operation / operating state of the mechanism unit and system associated with each process sequence. is there. The data group is continuous time series data. Here, if the number of data points from the start of measurement to the end of measurement is k, the number of processes (device operation etc.) performed in the process is ni, and the number of data is xi,
k = x1n1 + x2n2 +++ ini
And the number of sections is n.
(The data structure of x1, x2, x3,, xi consists of “photometer measurement value” and “apparatus operation / operation index value”. “Apparatus operation / operation index value” corresponds to the processing sequence. If necessary, add it, and if it is unnecessary, use only the “photometer measurement value”.)
This n corresponds to the reaction time and processing time of the process. The data structure is shown below.
(1) A standard Mahalanobis space is created from k pieces of data that have been measured successfully. In the k pieces of data, for example, the Mahalanobis distance itself, which is an index value indicating the state of the device that operates and operates in each process, is created.
(2) The k data of item (1) is divided into predetermined time intervals or every ni, and new Mahalanobis spaces are created in conjunction with the process progress. For example, when the operation / operation index value (●) of the device is added to the n1 and n4 areas, one of k3 and (k− (k1)) is the operation / operation index value. ing.

(3) The Mahalanobis distance is calculated at the end of the process measurement. For example, if n1 ends (t1), the data is addressed to the k3 reference spaces when the data collection is completed, and the Mahalanobis distance MD1 is calculated.
(4) By repeating the item (3) until the end of the measurement, the temporal change in the Mahalanobis distance for each category can be seen. That is, as shown in the table below, Mahalanobis distances MD1, Md2, MD3, and MD4 (for example, in the case of four divisions) are calculated, and the temporal change is known.

  The Mahalanobis distance that ended in the normal process is usually 0-4 (when the threshold is set to 4), so if any Mahalanobis distance in each process exceeds 4, some abnormality has occurred in the process. I understand. In the example of the above table, an abnormality has occurred in the sections n1, n3 and n4. Since the process is sequential, when focusing only on the n1 interval, the abnormality is mainly due to the higher contribution of either the “photometer measurement value” or “apparatus operation / operation index value”. I can judge. If the contribution of the “photometer measurement value” is higher, some other factors (for example, deterioration of the reaction vessel or It is determined that the measured value is abnormal due to dirt or the like.

Further, each reference space in each process has a Mahalanobis distance in the range of 0 to 4 and n processes as shown in the above table. For this reason, it is possible to make a more comprehensive judgment by storing the Mahalanobis distance calculated in each process until the completion of measurement and using each value as one comprehensive judgment data. (The reference space at this time is an aggregate of Mahalanobis distances taking into account the thresholds at the time of creating each reference space of n processes. In the above table, MD2 is normal in four sections, and is a Mahalanobis distance of 0-4. is there.)
In addition, a calculation unit that obtains measurement results that are different from the data measured by the automatic analyzer (inspection item) and have analysis conditions within a certain tolerance (matched) to build a database on reaction processes Is provided.

  In such a calculation unit, a basic data group for calculating the Mahalanobis distance is stored. The data group is an eigenvector (kxk) matrix calculated from an average value (k), standard deviation (k) and correlation coefficient matrix in each measurement time or a specific section, and the maximum required number of axes.

  As described above, such data has a strong correlation between the data. For this reason, the inverse matrix of the correlation coefficient matrix used in the process of calculating the Mahalanobis distance cannot be calculated. For this reason, in this method, an eigenvalue and an eigenvector of the correlation coefficient matrix are obtained, and the number of axes (mode) is determined based on the degree of contribution, and arithmetic processing for calculating the Mahalanobis distance is performed. In this calculation process, it is not necessary to obtain any inverse determinant from the correlation coefficient matrix, and the Mahalanobis distance can be calculated even if the correlation coefficient is “1.0”.

  Furthermore, an arithmetic processing unit for optimizing the determination logic is provided using the measurement results obtained by different results in the measurement under different conditions.

  As described above, the “apparatus operation / operation index value” is used as one of the verification data in each process, but it can be used alone as “apparatus operation / operation dynamic monitoring and management index”. Apply.

  This is because there is a limited-life product in the mechanism unit and subsystem included in the apparatus, and the mechanism is also deteriorated. For this reason, by tracking and monitoring / recording such a state change at a predetermined interval, information for effectively performing the maintenance plan of the apparatus can be created.

  The present invention has the following effects.

  Inspection reliability and device reliability are improved. In particular, even if an abnormality occurs in the reaction process, it is possible to detect an abnormal measurement result in which the measurement data falls within the normal value range. Even if the inspection result is out of the normal value range, if the reaction process data is not abnormal, it is measured correctly, so that unnecessary re-inspection is unnecessary.

  Furthermore, even when an abnormality is detected in the reaction process, it is possible to quickly determine whether or not the cause is an intrinsic factor in the apparatus or an extrinsic factor, thereby improving the inspection efficiency.

  Improved reliability of measurement items for specimens, reducing unnecessary retests and reducing running costs (ie consumables such as test reagents and cleaning fluids) for automated analyzer testing it can. In addition, the inspection time can be shortened.

  Further, since the reaction process data can be stored only for the measurement abnormality data, the cost for data storage can be suppressed.

  The present invention provides means for determining whether or not an inspection has been properly performed by using “measured values” and “apparatus operation / operation index values” in a reaction process, and provides thousands to several days a day. The purpose is to prevent oversight of items that show abnormal reactions even when 10,000 tests are measured.

  As such a method, a Mahalanobis space serving as a reference is formed from time-series data as in the above-described conventional example, the data is addressed to the space, and the determination is made based on the distance. However, as described above, in such an analyzer, the prozone / check error, sampling error, reagent dispensing error, stirring mechanism error, high-viscosity reagent dropout and scattering, and reagent combination caused by measurement data Nozzle contamination, crystal precipitation, etc. are composed of time series data. For this reason, an abnormality detection method that can cope with the time-series operation of each device is important.

  In the conventional method, since it is a comprehensive judgment based only on “measured values”, the reference space is generally configured using time series data from the start of measurement to the end of measurement. It is a comprehensive judgment. Therefore, it is difficult to determine which part of the analysis process was abnormal. Moreover, it was not possible to quickly determine whether the cause was an internal factor of the device (deterioration of the operating part, etc.) or an external factor (contamination, deterioration, etc. of parts of other mechanism parts, operation outside the specifications, etc.) . Furthermore, since this time series data is a data group in which the correlation between each data is strong (a constant value within the measurement error and the same gradient), it is necessary to devise a Mahalanobis distance calculation. Furthermore, it is necessary to devise in calculating an index suitable for discriminating the directionality of the Mahalanobis distance and the pattern difference of measurement items that are constituent elements thereof.

  Hereinafter, how the MT method was applied to the analysis of the reaction process will be described using examples. FIG. 1 shows a configuration example of the automatic analyzer. The main mechanism of the automatic analyzer is composed of a specimen disk 2, a reaction disk 1, and a reagent disk 3.

  Before starting the analysis process, several samples are installed in advance on the sample disk. When analysis is started, a predetermined amount of sample is aspirated by the sample dispensing mechanism (sample dispensing mechanism) 4 and discharged to a predetermined position on the reaction disk. For example, the specimen on the reaction disk is analyzed by a predetermined analysis sequence incorporated in the apparatus in advance as shown in FIG.

  FIG. 1 shows the operating position of each mechanism section with the reaction disk 1 as the center. A light source lamp 20 (see FIG. 2) for measuring the absorbance of the specimen is provided on the inner periphery of the reaction disk, and a photometer unit 7 is installed on the outer periphery thereof. Each time the reaction vessel 12 (see FIG. 2) on the reaction disk passes between the light source and the photometer, the absorbance is measured. The absorbance is measured after the reaction disk starts rotating and is accelerated to a constant speed. The reaction disk rotates and stops at a constant angle every cycle, and is measured many times during a predetermined reaction time.

  The control of these mechanical systems is mainly executed by a computer unit called a control unit 11, and an operation computer 15 for receiving sample information, reagent management information, examination request reception, and the like is connected. And is working.

The structure of the photometer unit 7 used in this embodiment is shown in FIG. The photometer unit used in this embodiment is called a post-spectral multiwavelength photometer. That is, the light emitted from the light source lamp 20 passes through the reaction vessel 12 containing the specimen, and then enters the concave diffraction grating 22 as a linear light beam through the entrance slit 21. Here, the luminous intensity of the light that has been split into multiple wavelengths and transmitted through the specimen is measured by a photometer 23 with 12 wavelengths.
1.1 Analysis Sequence FIG. 3 shows a flow chart of sample analysis performed in this example.

  Blood (white blood cells, etc.), cerebrospinal fluid, urine, or the like is used as the specimen, and is previously installed in one specimen container 13 (FIG. 1) on the specimen disc 2. This sample is dispensed into the reaction vessel 12 on the reaction disk 1 for analysis. As preparation before dispensing the specimen, the reaction container on the reaction disk is washed (A01), and a water blank is measured (A02). The water blank is to measure the absorbance of water in order to adjust the specimen absorbance to zero. That is, the absorbance value of the sample dispensed in the reaction container is obtained by the difference from the absorbance value of the water blank. When the measurement of the water blank is completed, the water in the reaction vessel is sucked and discarded (A03). A predetermined specimen is dispensed (sampled) into the reaction container (A04). Thereafter, R1 reagent (A05), R2 (A07) reagent, R3 reagent (A09), and R4 reagent (A11) are added to the reaction vessel in a predetermined amount at a predetermined time, and stirring (A06, A08, A10) is performed. , A12). Here, depending on the analysis item, there is an inspection item in which R4, R3, or R2 is not dispensed. In general, the reaction process includes a 3-minute reaction, a 4-minute reaction, a 5-minute reaction, and a 10-minute reaction, and the absorbance when each reaction disk rotates by the number of times corresponding to the reaction time is used as a measured value. Usually, the reaction is often carried out for 10 minutes. When a predetermined reaction time has elapsed and all photometry is completed (A13), the reaction vessel is washed for the next analysis (A01).

Description of Concentration Calculated Absorbance FIG. 4 shows a typical method for analyzing the absorbance thus obtained. For the concentration calculation from the absorbance, a 1-point analysis method, a 2-point rate analysis method, a 2-point analysis method, a 3-point 2-item analysis method, or the like is used. In the one-point analysis method, the concentration of the component to be examined is calculated from the absorbance after a certain time has elapsed since the addition of the reagent. In the two-point rate analysis method, the difference in absorbance at two times t1 and t2 (t2> t1) determined from reagent addition is expressed as (t2−
The concentration is calculated from the time ratio of absorbance change divided by t1). In the two-point analysis method, the absorbance at two times t1 and t2 (t2> t1) determined from the reagent addition is measured, and the absorbance at t1 is subtracted from the absorbance at t2 multiplied by a constant factor. The concentration is calculated.

  In any case, the absorbance is measured every time the specimen on the reaction disk crosses the photometer, and the concentration of the test target component is determined by arithmetic processing using a part of the measured value. That is, most (or part) of the absorbance measured in the reaction process has been discarded in the conventional analysis method.

Reaction Process Data Abnormality Detection Algorithm FIGS. 5 and 6 are diagrams showing the overall processing flow of reaction process abnormality detection according to the present invention and the details thereof.

  First, in order to detect an abnormality in each process, data collection is performed in accordance with a predetermined inspection / measurement sequence for each process.

For example, in the reaction process of A01 to A02 shown in FIG. 3, a set of reaction process data (output data group of the photometer 23) for the test to be determined first is taken into the computer in order to detect abnormality of the process ( S1). At this time, in addition to the output value of the photometer 23, the ambient temperature at which this system is operating and the value of an amplification circuit (not shown) for amplifying and standardizing the output of the photometer 23 are taken in (S41). ). In the process, the absorbance of water is measured in order to adjust the absorbance of the specimen to zero. In the reaction container 12 (FIG. 2), the absorbance value of the sample dispensed in the subsequent processing sequence (FIG. 3) is measured, and this absorbance value is obtained by the difference from the absorbance value of the water blank. The reproducibility and reliability of the system including the photometer 23 in the process (measurement system including the light source lamp 20, the entrance slit 21, and the concave diffraction grating 22) is important. However, when the reaction vessel 12 is deteriorated or contaminated, when the reaction vessel is not sufficiently cleaned (A0 in FIG. 3), when the operating temperature or the ambient temperature changes for some reason, When the deterioration of other electrical components or the like accelerates rapidly, the absorbance value of water, which is the measurement value, changes, and the absorbance value of the dispensed sample also changes. For this reason, the entire process of A01 to A02 is verified by adding an index value indicating the operating state of the system and a change in state to the output value of the photometer 23.

  The timing of capturing the reaction process group data matches the predetermined measurement sequence for each process. Alternatively, absorbance data that is sequentially measured in the reaction process may be taken in and a series of time-series data groups may be obtained later.

  In addition, the above-described system operation state and state change index values are sequentially added corresponding to the mechanism units and subsystems in the reaction process associated with the measurement sequence shown in FIG. The index value of the operation state or state change is, for example, a Mahalanobis distance calculated from a large number of measurement items as shown in FIGS. 5 and 6 (S42).

  Next, the Mahalanobis distance is calculated | required about the reaction process data obtained from the above (S2). The data at this time is composed of an output value data group of the photometer 23 and index value data (Mahalanobis distance) that comprehensively verifies the operation status and state change of the system including the photometer 23.

  Here, the Mahalanobis distance is a method of multivariate analysis, and it is a scale for measuring whether a certain test object belongs to a reference group (hereinafter referred to as a reference space).

  In the present invention, a reference space is formed from a group of reaction process data when analysis is normally performed, and the information is stored in a database DB as shown in FIGS. How to obtain the Mahalanobis distance and the reference space will be described later.

  Next, it was collected from the Mahalanobis distance calculated from the output value data group of the photometer 23 and the index value data (Mahalanobis distance) that comprehensively verified the operating status and state change of the system including the photometer 23. Determine whether the data group (subject sample) is abnormal or normal. The Mahalanobis distance when the dispensing is carried out normally shows one value, whereas when it is abnormal, the distance takes an extremely larger value than 1.0 (average value). Utilizing this, determination of normality / abnormality of dispensing is performed more comprehensively by threshold determination (S3).

  On the other hand, the index value data (Mahalanobis distance) that comprehensively verifies the operating status and state change of the system including the photometer 23 indicates the temporal operating status change and status change of the system. As described above, it is also possible to determine and determine whether an abnormality is normal or normal. While the Mahalanobis distance is 1.0 (average value) when the system is operating and operating normally, external factors other than the system, such as contamination of the container 12 and rapid progress of parts with a limited life, etc. ) Occurs, the distance is much larger than 1.0 (average value). Utilizing this, the normality / abnormality of the mechanism or system corresponding to the reaction process is determined (S43) and the progress is monitored (S44). In the final determination, the overall determination value s3 may be matched with the determination result value s43 of the mechanism unit.

  Regarding the reaction process data in this embodiment, the data structure and the reference space will be described with reference to FIGS.

  FIG. 7 shows a method for obtaining the reaction process data A01 to A02 shown in FIG. 3 for one wavelength. Each time the specimen sample passes through the photometric point, the absorbance of the specimen is measured and sequentially accumulated in a computer (see FIG. 6). Finally, the predetermined number k (time series) as shown in FIG. ) Is taken in, and n absorbances (number of samples) are taken in (S1). At this time, as shown in FIG. 7, the operating state and state change of the system including the photometer 23 are determined in advance in synchronization with the measurement of the absorbance of the specimen every time the specimen sample passes the photometric point. The output values (photocurrents) of the 12 wavelengths of the photometer 23, the ambient temperature at which this system is operating, the conversion circuit constants, and the like are captured. The Mahalanobis distance s42 is calculated from the ks pieces of data, and the calculated Mahalanobis distance “md11” is finally added to k pieces of data (time series) determined in advance as shown in FIG. One piece of absorbance (number of samples) data is taken in (S1).

  Among the k data, in this embodiment, the first four points are the data of the A01 to A02 processes, and three of them are the absorbance of the cell blank value from the photometer 23, that is, the zero point of the absorbance. It is. On the other hand, the other one is the value of Mahalanobis distance “md11”: s42 which is an index value of the operation / operation state of the system including the photometer 23. In this example, the fifth and subsequent samples are a data group (reaction process) obtained by tracing the temporal change in absorbance after the introduction of the reagent from the suction of water, but in the fourth section, as described above, in the final reaction process. The operation / operation index value “md12”: s42 ′ value of the mechanism unit or subsystem.

  The number of measurement points for the cell blank value is determined by the mechanism system and the control method, and in the actual analysis, for example, the absorbance for multiple wavelengths such as 12 wavelengths is measured. The absorbance of the second wavelength is changed from the (k + 1) th to the 2kth, and similarly, the absorbance of the twelfth wavelength is changed from the (11k + 1) th to the 12kth, and 12k absorbance data are acquired. By using the absorbance for all or a part of the wavelengths, it is possible to detect the reaction process abnormality with higher accuracy. However, for the sake of simplicity in this example, the measurement value of the cell blank value is 1 Only the wavelength is described.

  The photometric points for each wavelength are measured at regular intervals, but the absorbance data may be increased by installing multiple photometers, and the absorbance at many photometric points where reaction process abnormalities are likely to occur. On the other hand, the absorbance data may be thinned and fetched at a location where there is almost no abnormality, and it is not always necessary to have the same interval.

  In the present embodiment, as described above, the blank process and the reaction process are roughly divided into two, and the process in the middle is divided into two and finally divided into four sections. May be divided in more detail in accordance with Furthermore, the index values “Md11” and “Md12” of the operation and operation of the mechanism and subsystem corresponding to the above-described reaction process can be arbitrarily added corresponding to each reaction process, and are not adopted unless necessary. As good.

  The reaction process data acquired as described above (in this embodiment, four processes are included) are summarized as shown in FIG. Here, the absorbance at each photometric point and the index value of operation / operation are used as items for obtaining the Mahalanobis distance.

  If the above operation is performed on a reaction process data group for which the measurement result is determined to have been completed normally, a reference reaction process data group can be obtained. FIG. 8 summarizes each reaction process data group obtained when normal dispensing is performed n times. This is an n event k item, and an index of operation / operation in the first and fourth zones. The reference space includes one value.

  The n normal reaction process data groups referred to here are reaction processes when a reproducible result is obtained for a sample whose measurement result is within a range in which accuracy is guaranteed in the automatic analyzer. In addition, this is a reaction process when the operation / operation state of the mechanism unit and the subsystem in each process is normal.

  For example, if abnormalities such as the prozone phenomenon or sample / reagent dispensing occur, these are due to accidental phenomena. Even if it is normal, there is no reproducibility. That is, even if reproducible measurement is performed by changing the apparatus, time, reagent and sample dispensing amount, etc., the same result cannot be obtained, so this is not a normal reaction process.

  In addition, since n data groups are used and a reference space is created from these data groups, it is not just the collection of statistical data, but the collection of statistical data when normal as described above. There must be no abnormal data. However, there are variations within the range in which the instrument guarantees measurement accuracy, such as variations in specimen characteristics (measurement values), variations in photometers, and variations in the allowable operation of each mechanism and subsystem. If this is the case, it is desirable to actively disperse and obtain data, and by doing so, the accuracy of abnormality detection is improved.

  As for the number n of events in the reference space, a larger number is preferable because the accuracy of abnormality detection is improved. However, if the number is larger than necessary, the economic cost is higher than the obtained information. Therefore, it is preferable to make a decision in consideration of detection accuracy and economic cost. However, when the number of events n is smaller than the number of items k, a correlation matrix to be described later cannot be obtained, so that n must be larger than k.

  It should be noted that the reaction process data group and the reference space obtained as described above are limited to the analysis part and the mechanism part or subsystem in each reaction process corresponding to the analysis item, and the reaction process data group and the reference space for each analysis item. A reference space must be prepared. This is because, depending on the reaction item, the analysis time, the operation of the mechanism part and subsystem, the type and amount of the reagent used vary, and the time change pattern of absorbance varies greatly.

Details of the method of calculating the Mahalanobis space (reference space) for the reaction process data p ₁ , p ₂ ,..., P _k as shown in FIG.

  Furthermore, FIG. 9 shows a calculation flow of the Mahalanobis distance at the time of data collection input of the detected object.

About the reaction process data composing the standard space of n events k items (photometry points) as shown in Fig. 8, the average value for each item (photometry points and index values of operation / operation)

Then, standard deviations σ ₁ , σ ₂ ,..., Σ _k are obtained, and normalized by performing the calculation of equation (1).

On the other hand, when the reference space is a matrix of n rows and k columns, and a correlation matrix of this matrix is obtained, a k × k matrix A is obtained. If the inverse matrix of the matrix A is A ⁻¹ , the Mahalanobis distance D ² can be expressed as in equation (2).

  Of these calculations, the average and standard deviation for each item of the reference space (photometric point operation / operation index value) and the inverse matrix of the correlation matrix of the reference space are calculated in advance. The result is included as a parameter (see FIGS. 5 and 10). When the Mahalanobis distance is repeatedly calculated, it is possible to save time and effort for performing these calculations many times, which is advantageous in calculation processing.

Here, if the reaction process data p ₁ , p ₂ ,..., P _k are normal, the Mahalanobis distance D ² is a value near 1.0, and if it is abnormal, it is generally compared with 1.0. Take a large value. In order to discriminate between normal and abnormal based on the Mahalanobis distance D ² , a certain threshold value x is set in advance,
Normal if D ² <x,
If D ² ≧ x, it is determined as abnormal.

  In the time series data group shown in FIG. 8 described above, there are many cases where the correlation between the data is strong due to the nature thereof, and there is a case where the inverse matrix is not obtained in the process of calculating the above equation (2). To do.

  As a countermeasure in such a case, in general, any one of items with strong correlation may be deleted sequentially. However, in this method and its processing method, further data processing and its processing task must be additionally installed, so that its processing capability is reduced and its reliability is also reduced due to a reduction in the number of data points. For this reason, in the present invention, a processing task is devised and applied without any further processing of the collected data group even when the correlation between the data is strong. Hereinafter, the contents will be described in detail.

As described above, as a method of calculating the Mahalanobis distance, the method according to the above equation (2), the Schmitt orthogonal expansion method (document: technical development in MT system), and the principal component analysis are used to orthogonalize the data group and to create a correlation matrix. Various calculation methods that apply a cofactor matrix that does not obtain an inverse matrix in the first place and a processing method that singularly decomposes s. Each method has its characteristics and is still developing, and it takes time to apply it to a system that emphasizes real-time performance as in the present invention. Therefore, in the present invention, a new Mahalanobis distance processing method based on the principal component analysis type has been found based on various experiments and verification results.
Hereinafter, the outline of the processing flow will be described below with reference to FIG.

A correlation matrix (k * k matrix) and eigenvector are obtained. (K is the number of items)
The contribution rate of the eigenvalue and the required number kn are calculated. (Maximum k)
From the result of the term (2), if kn <k, only kn and kn + 1 are used, and the Mahalanobis distance (MD2) is calculated from the eigenvector obtained in the term (1).

  If kn = k, k pieces are used, and the Mahalanobis distance calculation is calculated from the eigenvector obtained in the item (1). Alternatively, assuming that kn = k, the knth and subsequent eigenvalues at which the contribution rate is 1.0 are changed within the error range, and the Mahalanobis distance calculation is calculated from the eigenvectors obtained in the item (1).

Furthermore, in the present invention, two vectors are easily calculated in the calculation process of the above item (3) ((4) -1 and (4) -2 in FIG. 9). One is a vector (yva) for calculating the Mahalanobis distance, and the other is a vector (xva) orthogonal thereto. Here, as shown in FIG. 9, the Mahalanobis distance (MD2) is Md ² = Σyva (i) ^ 2 / kn.
((5) in FIG. 9). (I = 1 to kn, kn is the number of measurement items)
On the other hand, xva (i) indicates the behavior of the measurement item in the vertical direction (orthogonal direction) at the horizontal position (yva (i)) of the measurement item that determines the Mahalanobis distance. Therefore, Σ {yva (i) * Σxva (i)} indicates the overall behavior of the entire measurement item, and for each measurement item of the target event with respect to the behavior of each measurement item when creating the reference space. It becomes an index to inform the contribution degree and the overall behavior difference (pattern difference).

  For this reason, for example, when the number of measurement items is limited due to hardware limitations, the degree of contribution was examined and selected with the above-mentioned index, and the sensitivity and variation of the Mahalanobis distance were maintained. This makes it possible to reduce the number of items.

  Furthermore, for example, since the reference space for the above-mentioned index values for operation / operation is generally formed at the time of installation of the device or on the operation start date, as the operation / operation time of the device increases, Mahalanobis distance also shows a monotonous increase. However, there may be cases where the average operation / operation state (for example, when the effective continuous use period is 5 years, during the operation period of 2.5 years) is used as the reference space for monitoring / management of the operation status of the equipment and maintenance plan. is there. In such a case, if the reference space is “0 point” and a part with a limited life is replaced, the distance will change in a favorable direction at the time of replacement, and this “0 point” will change as time passes. Pass through and get worse. Even in such a case, the Mahalanobis distance can be given directionality by considering the index of Σ {yva (i) * Σxva (i)}. The contents of the index will be described later.

  In such a process, there is no additional processing of the collected data group. Even if there is a strong correlation between items, the eigenvalues of the correlation matrix are calculated first, so k eigenvalues (orthogonal axes) and their contributions are calculated. Since it is possible to determine which number of axes should be applied based on the degree of contribution, unnecessary axes can be easily deleted, and errors in calculating the Mahalanobis distance can be minimized. Furthermore, even if there is a strong correlation between the items with a correlation coefficient of “1.0” (exactly the same data), it is confirmed that the Mahalanobis distance can be calculated without any data processing of deleting one item. did.

  Furthermore, this processing (the above calculation process) and the above equation (2) are performed using data in which the correlation coefficient between items is relatively small and the inverse matrix can be obtained without deleting the item in equation (2). As a result of comparison with the result calculated by (3), the calculated Mahalanobis distance was not significantly different (within the numerical error). The validity and reliability of this process are confirmed.

  FIG. 10 is a functional block diagram incorporating the discrimination logic of the reaction process of FIG. The analysis control unit 31 is a function implemented on the control unit (control computer unit 11), and other functions and data are implemented on the operation computer 15.

  The analysis request receiving unit 30 is for setting what kind of analysis test the operator performs on the sample, and is performed using a screen such as a CRT and an input device such as a keyboard and a mouse. A control command is sent to the analysis control unit 31 from the input information. The analysis control unit 31 executes the analysis by controlling the mechanism shown in FIG. 1, and dispenses the sample on the sample disk 2 onto the reaction disk 1 to perform the reaction. When the analysis is completed for one sample, the reaction process data 34 and the analysis result data 32 at that time are stored in the database. Whether or not the reaction process data 34 is abnormal is determined by the reaction process evaluation unit 35. At this time, the evaluation is performed with reference to the reference space data 33. This reference space is prepared for each analysis item, and the reaction process evaluation unit further includes analysis items and mechanisms and subsystems corresponding to the reaction process. Refer only to the reference space corresponding to the operation / operation.

  When the reaction process evaluation unit 35 determines that the reaction process is abnormal, the analysis control unit 31 is instructed to reexamine the same test item for the same sample. Alternatively, if the contribution of abnormalities in the operation / operation of the mechanism unit or subsystem in the process is large, information to that effect is also added. Further, information indicating that there is an abnormality in the reaction process is added to the stored analysis result data 32.

  In the device maintenance information unit 36, the measurement value information for each process of the device (process) when the reaction process is determined to be normal or abnormal in the reaction process evaluation unit 35, and the mechanism unit that operates corresponding to each process, Subsystem index values are stored as data in time series.

  With such information, for example, the Mahalanobis distance calculated from data collected in a certain process and the elapsed operation time of the device are monitored, and the temporal transition and deterioration status due to internal factors such as component deterioration inside the device, Alternatively, it can be monitored whether it is caused by a sudden external factor. These trend data are useful as maintenance information for the device itself.

  FIG. 11-A (at the time of creating the reference space) shows an example of a data group (sample number: 110 items, item number: 30 items) of the reaction process that has been normally completed.

FIG. 11-B shows an example of a data group (number of samples: 34 samples) collected from the detected object separately from the normal reaction process data group.

  FIG. 12 shows an example of a typical result when a reference space is created from the data group of FIG. 11-A and each data of FIG. 11-B is addressed to the reference space (s-5, s-7, s-13, s-18, s-24, s-29) are shown for each category corresponding to each reaction process (time course of reaction process). In this example, as described above, the section is set to four sections, and the threshold value in the section is set to 4.5 in each section.

  The s-24 and s-29 of the data of the detected object are determined to be “normal end” because the Mahalanobis distance is within the threshold regardless of the passage of time (progress of reaction process) in any section. The

  On the other hand, s-18, s-13, and s-5 exceed the threshold at the stage of the initial division (initial reaction process region), and the Mahalanobis distance value tends to increase with time. This is because some abnormality has occurred in the process (process) of section 1, and the influence continues throughout the entire reaction process, and it can be determined that the process is “abnormal end”. Alternatively, an “abnormal” alarm is generated at the end of the measurement in the section 1, and a procedure for not measuring or preparing a new sample separately, or a sequence change can be automatically performed.

  At this time, as described above, the device maintenance information unit 36 responds to the measurement value information for each process of the device and the process when the reaction process evaluation unit 35 determines that the reaction process is normal or abnormal. The index values of the mechanisms and subsystems that have been operated are stored as data in time series. An example of the disclosure result is shown in FIG.

  The index value of the operating / operating state of the system including the photometer 23 in this process (first section) of the specimen is approximately 4.0 (circle P portion in the figure), and the Mahalanobis distance space when creating the reference space Was located within. However, among the system information including the photometer 23 (S41 in FIG. 7), the ambient temperature and the conversion circuit constants were within the allowable range, but only λ3, λ4, and λ6 within the 12 wavelengths were near the limit of the allowable value. there were. From this index value, it can be judged that the main factor is highly likely to be an external factor due to deterioration or contamination of the reaction vessel 12 used in this process. It is possible to perform processing such as promptly giving instructions.

  As described above, abnormalities such as sampling, reagent dispensing, and stirring may occur as the cause of the abnormality in the reaction process. For example, the reagent discharged during the addition of the R2 reagent may adhere to the side wall of the reaction cell and mix with the sample in the middle of the reaction process, so the result may be higher than the true value. In addition, when a reagent having a higher viscosity is used, the discharged reagent remains as a water droplet at the tip of the dispensing nozzle due to surface tension, the nozzle is contaminated by a combination of reagents, and crystal precipitation occurs.

  Normally, the measurement result (concentration value) in one analytical test is only one real number, and when the measured value in a certain range is obtained, a function such as automatically performing a retest is implemented in the analyzer. Reproducibility is confirmed. If there is an abnormality in the measured value, it may be possible to determine whether it is abnormal or not by examining the absorbance in the reaction process. However, in the system configuration with 50 photometry points, all the reaction process data of 12 wavelengths are stored. It was necessary to store 600 pieces of data in one measurement. For this reason, even if there is an abnormality in the reaction process, if the measured value falls within the normal range, the abnormality may be overlooked. Furthermore, since there was no information that could analyze these main factors, the anomaly could be overlooked.

  The reaction process in which an abnormality is detected from the reaction process data often includes information for determining what is the cause of the abnormality by analyzing the reaction process data. Therefore, if the reaction process data is saved only when an abnormality is detected in the reaction process data, the memory capacity for storage such as a hard disk can be reduced, and once an abnormality is detected, the reaction process data can be used. The cause can be investigated. The cause of the abnormality in the reaction process data may be an abnormality in the analyzer itself. Therefore, by collecting information on the mechanism part of the device, subsystem information, and control system information that accompanies each reaction process, Phenomenon analysis is performed quickly.

  As described above, in the reaction process evaluation unit 35, when an abnormality occurs in the reaction process, the presence / absence in the process and the change in the operation / operation state of the mechanism unit or subsystem of the apparatus in the process are notified. This contributes to the analysis of abnormal phenomena and the speed of post-processing.

  As described above, the Mahalanobis distance is used as an indicator for determining whether there is a change in the state or whether there is an abnormality. However, in the present invention, a new contrivance is implemented in order to further speed up the analysis of the abnormal phenomenon. The contents will be described below.

  FIG. 14 shows the positions of points A, B, C and D and their Mahalanobis distance in the case of two variables (x1, x2). As shown in the figure, the Mahalanobis distances of the points A, B, C, and D are exactly the same, and their appearance probabilities are also exactly the same. However, their positions are different. In other words, even if the Mahalanobis distance is the same, the arrangement pattern of the coordinates (x1, x2) is different. Therefore, if a judgment is made in consideration of this pattern difference, a judgment with higher accuracy can be made. However, in the conventional evaluation and discrimination using the Mahalanobis distance, a suitable index that can cope with this kind of problem (Mahalanobis distance directionality, pattern difference identification of measurement items) has not been obtained.

  In the present invention, as shown in FIG. 14, feature quantities 1 and 2 are separately calculated for the points A, B, C, and D, and a code is added to the Mahalanobis distance by a combination of the codes. Further, the contribution amount of the measurement item can be determined from the feature amount 3.

The calculation process is shown in FIG. In the present invention, in the process of calculating the Mahalanobis distance (see FIG. 9), not only the vector Yva ((4) -1) for calculating the Mahalanobis distance, but also the vector Xva that is an orthogonal component thereof is calculated simultaneously. . The vector: Yva is composed of k measurement items, and each point indicates the position (pattern determination position) of the total behavior of the measurement item for which the Mahalanobis distance is determined. On the other hand, Xva, which is an orthogonal component of Yva, indicates the behavior of each measurement item (variation value for each item) itself. From these two vectors, feature quantities that enable determination and identification of the directionality of the Mahalanobis distance and the pattern difference of the measurement item are calculated.
Feature amount and identification for calculating directionality of Mahalanobis distance Feature amount 1: (6) -1 in FIG.
From the Yva (see FIG. 9-2), the total value of the fluctuating energy taking the directionality (+, −) into consideration is calculated. At this time, if the sum is + (positive), it is assumed that this event (for example, points A and B) is biased toward the “+” side with respect to the reference event.

Feature amount 2: (6) -2 in FIG.
The total value of the sum of products of the position of the total behavior of the measurement item that determines the Mahalanobis distance (pattern determination position: Yva) and the behavior of each measurement item (variation value for each item: Xva) is calculated. At this time, if the total value is “+”, it is assumed that this event (for example, A, C) is biased toward the “+” side with respect to the reference event. As described above, this index value is the product-sum of the position of the total behavior of the measurement item and the behavior of each measurement item at that point, so it suggests more comprehensively the energy fluctuation and directionality of each measurement item. Yes. Such an index is an index found from various experiments and verification results.

The conventional Mahalanobis distance and the Mahalanobis distance calculated by (5) of FIG. 9 described above are decomposed into distances having four directions from the feature quantities of the items (1) and (2). (The results of this example are shown in FIG. 16-1)
Features for identifying pattern differences of measurement items: (7) -1 and (7) -2 in FIG.
Among the above-mentioned items (1), in the calculation process of the feature amount 2, for each sample, the position of the total behavior of each measurement item and the product sum of the behavior of each measurement item at that point are calculated for each measurement item. . That is, since the calculated value for each measurement item indicates the overall behavior for each measurement item, the pattern difference or contribution is determined by comparing or ranking the behavior. I can judge. In other words, a large variation of each measurement item at the time of creating the reference space has a large contribution to the calculated Mahalanobis distance, which is a main factor causing a pattern difference. On the other hand, if it is within the fluctuation at the time of creating the reference space or ≈0, the contribution is small and it is not the main factor causing the pattern difference ((7) -1).

Furthermore, in order to eliminate the influence of sudden fluctuations (events in which one of the k items is moving in the opposite direction), the value of the above-mentioned value ((7) -1) (generated from actually measured data, not estimated values) The weight is given to the distribution value (σ (i) ² ) of each item and the dispersion value (Σσ (i) ² ) of all items to obtain an evaluation index. (The results of this example are shown in FIG. 16-2)
FIG. 16 shows the result of recalculation of the data of the reaction process exemplified in FIG. 11 by the above calculation. Sample S-33 in the table is data at the time of creation of the reference space, and its value is a sample in the vicinity of the average, and its Mahalanobis distance (Mdx2 column) is ≈0. On the other hand, the Mahalanobis distances of samples S-23 and S-24 are 86.5 and 87.5, and the Mahalanobis distances of samples S-25 and S-26 are abnormal data of 346.4 and 348.9. From this distance, it can be seen that the degree of abnormality in S-25 and S-26 is larger than that in S-23 and S-24, and therefore the degree of abnormality is large. However, the contents and the causal system are unknown in the past, but according to this method, the signs are different as shown in the column “direction” in the table and as shown in FIG. It is clearly divided into four indicators.

  That is, although the Mahalanobis distance is the same in S-23 and S-24, the behavior pattern of the measurement item is different, and S-23 is on the “+” side with respect to the item variation at the time of creating the reference space. It is abnormal, and it can be determined and identified that S-24 exhibits a pattern of a position that is approximately 90 degrees in phase with respect to S-23.

  Furthermore, as shown in FIG. 16-2, when the variation pattern for each item is seen according to the feature amount 3 described above, it is clear that only the behaviors of the items No. 19 and No. 21 are inverted. It can be determined that an abnormality in the operation of the system related to the measurement items No. 19 and No. 21 or an abnormality in the measurement system or reaction process has occurred.

  On the other hand, the sample S-25 has the same directionality as the sample S-23, but the operation of the system related to the measurement items No. 19 and No. 21 is increased by the increase of the Mahalanobis distance or It can be determined that some degree of abnormality in the measurement system or reaction process is large (the main cause of the cause system is the same, but the degree of abnormality is more advanced).

  FIG. 17 shows an example in which the contribution level of an item using the above-described feature amount 3 (FIG. 16 (7) -1, (7) -2) is selected and determined with respect to the reaction process data of FIG. Show. In the figure, the upper figure shows the overall result, and the lower figure shows an enlarged scale.

  In addition, originally, such an operation or treatment is mainly used to quickly find a pattern difference between n groups of samples having the same Mahalanobis distance and the same directionality. However, in this example, the case where all samples are targeted is illustrated. This is the case, for example, when the number of measurement items is reduced due to hardware limitations or when the number of sensors is limited. Even in such cases, measurement is performed to prevent the sensitivity and variation of the Mahalanobis distance from decreasing. This is because it is often necessary to determine and select the number of items.

FIG. 17 shows the results for each item (30 items) calculated by the above (7) -1 and (7) -2. Note that the variation value of each item is indicated by its magnification (response magnification) based on the variation value of each item when the reference space is created.

  As is apparent from the figure, it can be seen that the response magnification of the four items (items No. 3, 24, 25) is not changed. Therefore, it can be determined that the item No. is not a factor that causes a pattern difference but is an unnecessary item.

  18 shows a comparison between the Mahalanobis distance when the items No. 3, 24, and 25 are deleted from the result of FIG. 17 and the calculation result of the Mahalanobis distance when the items are not deleted.

As shown in the figure, the correlation coefficient of the Mahalanobis distance when the item is deleted and when it is not deleted is 0.9995, and it is clear that there is no problem even if the item is deleted. In other words, we were able to reconfirm the validity and effectiveness of this indicator.

  As described above in detail using specific examples, according to the index according to the present invention, even when an abnormality occurs, the cause of the Mahalanobis distance can be quickly identified and identified, so that the analysis of the abnormal phenomenon And speeding up post-processing. Furthermore, since the accuracy of the value of the Mahalanobis distance is improved, it can be sufficiently used for maintaining and improving the reliability of the apparatus.

  In the device maintenance information section 36 (FIG. 10), the measurement value information for each process (process) of the apparatus when the reaction process is determined to be normal or abnormal, and the mechanism section that operates corresponding to each process, The index values of the subsystem are stored as data in time series, so by monitoring temporal changes in the Mahalanobis distance calculated from the collected data, it is applied as one piece of maintenance information for the equipment itself. It was. FIG. 19 shows an embodiment for more comprehensively using the above-mentioned maintenance information of the device itself. The difference from the description of the above-described embodiment (FIGS. 5 to 8) is that the operation / operation status of the mechanism unit and subsystem associated with each reaction process provided in the apparatus, the state change thereof, etc. Newly, it can be integrated and monitored.

  The various mechanical parts and subsystems (including the mechanical part and subsystems associated with the above-described reaction process part) included in the apparatus also have various life-long parts and wear and deterioration. The recommended operating period is also different. In the unlikely event of a sudden failure within the recommended operating period, parts may need to be replaced. In addition, at the time of periodic inspection of the apparatus, the above-mentioned limited-life parts and deteriorated parts are replaced. For this reason, for the purpose of ensuring and maintaining the reliability of the apparatus, it is important to more comprehensively monitor the operating part of the apparatus and track and disclose the transition of the operation status.

  Accordingly, as shown in FIG. 19, the s41 data of the mechanism unit and subsystem and the Mahalanobis distance s42 data associated with the reaction process separately provided in the apparatus are taken in simultaneously with other mechanism units and subsystems: S51, S61 data and Mahalanobis distances S52 and S62 data are collected and stored in the apparatus maintenance data 36 (see FIG. 10) in a more integrated form as shown in format s7. By calculating the Mahalanobis distance for such data: s7 periodically or at predetermined time intervals and disclosing or storing the result, it is possible to perform more comprehensive monitoring as a part of the entire apparatus. The disclosed example is shown in FIG.

In the disclosure example shown in FIG. 19A, the reference value (threshold value) for the determination based on the Mahalanobis distance is set to “average range of 0 to −2 to +2. The reference space of the disclosed example is the initial operation time of the apparatus. It is not based on the standard space created during periodic inspections, but is formed from data groups of each mechanism part that has been operating for a certain period of time, that is, from the average recommended period of each mechanism part and components with a limited life span and data on the deterioration state. Since the reference space is formed, if each mechanism section is more normal, a value on the − side is shown. On the other hand, if it is worse, a value on the + side is shown. (Addition of a sign is optional according to FIG. 15 and may be reversed from “+” to “−”). If there is a sudden abnormality in any of them, the value If it shows a sudden “+” value, or if the recommended period is exceeded, the value indicates a monotonous increase toward the “+” side. It is clear from the above description that this is possible.

  According to this data processing flow and the application of the above-mentioned evaluation index, the effect can be confirmed at the time of parts replacement at regular inspections, calibration of operating parts, or parts with a limited life, Additional work such as changing or reconfiguring the reference space again at the time of calibration can be deleted (unnecessary). Therefore, it is possible to continuously obtain lifetime data from the start date of operation of the apparatus until it is discarded, and it can be greatly useful as maintenance information of the apparatus itself.

  As described above, the embodiment according to the present invention has been described in detail. However, the operation index data (S42 in FIGS. 7 and 8) of the mechanism unit or subsystem associated with each reaction process described above is associated with each reaction process. It is clear that any mechanism or subsystem can be added or deleted arbitrarily in the reaction process.

  For example, in the embodiment of FIG. 3, since the dispensing mechanism 4 (see FIG. 1) for dispensing the sample is additionally provided in the process of A04: “sample sampling”, the operation / operation index is S42. Can be added in a similar format. Further, it can be added as one of the monitoring data shown in FIG.

  Furthermore, some cases have been disclosed [Non-Patent Document 2] in which, in the past, such data addition / deletion is performed again to rearrange those data and rearrange them in order of high contribution (Non-Patent Document 2). By adopting the feature quantity calculated by FIG. 9 “Mahalanobis distance calculation flow” and (6) -1, (6) -2, (7) -1, (7) -2 shown in FIG. Obviously, no additional work or treatment is required, and no indication of the directionality of the distance or treatment for changing the feature quantity is required.

Schematic mechanism diagram of an automatic analyzer. Configuration of post-spectral multiwavelength photometer and output example of photometer. Analysis flow. Example of reaction time course. Example of overall processing flow. Detailed example of processing flow. The structural example of reaction process data S1, S2, S41. Overall structure of reaction process data and its classification example. Mahalanobis distance calculation flow. Data flow diagram of the system. Example data of reaction process. An example of an abnormal reaction appearing in the measured data. Example of time transition of Mahalanobis distance of mechanical part or system. The figure explaining the directionality of Mahalanobis distance, and its feature-value. The flowchart which calculates the directionality of Mahalanobis distance, and its feature-value. Example of Mahalanobis distance direction and item pattern difference. Example of pattern difference (response value, contribution) for each measurement item. Comparison of Mahalanobis distance when the number of measurement items is reduced. A data flow for comprehensively monitoring and tracking a plurality of mechanical units or other subsystems in the apparatus, and a disclosed example thereof.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Reaction disk, 2 ... Specimen disk, 3 ... Reagent disk, 4 ... Sample dispensing mechanism, 5 ... Reagent dispensing mechanism, 6a ... Specimen dispensing nozzle washing part, 6b ... Reagent dispensing nozzle washing part, 7 ... Luminous intensity Total unit, 11 ... control unit (control computer unit), 12 ... reaction container, 13 ... sample container, 14 ... reagent container, 15 ... operation computer, 20 ... light source lamp, 21 ... slit, 22 ... concave diffraction grating, DESCRIPTION OF SYMBOLS 23 ... Multi-wavelength photometer, 30 ... Analysis request reception part, 31 ... Analysis control part, 32 ... Analysis result data, 33 ... Reference | standard space database, 34 ... Reaction process data, 35 ... Reaction process evaluation part, 36 ... Apparatus maintenance data .

Claims (10)

Measurement of the terminated reaction process normally, in the step of creating a Mahalanobis space as a reference a time series Retsude over data of the measurement values of the photometer, for each of a plurality of processes the time-series data determined in advance of the reaction process Dividing and creating a plurality of new Mahalanobis spaces from the time series data of the reaction process up to the time of the division ,
Calculating the Mahalanobis distance from the corresponding new Mahalanobis space for each of the time points from the time series data of the newly measured photometer ,
Determining the presence or absence of an abnormality based on the Mahalanobis distance for each of the plurality of predetermined processes;
A method for determining whether there is an abnormality in a measurement reaction process including:
Step, measurement results of the reaction process determined as normal, the time series data of the measurement values of the photometer, the operation-operation of the mechanism or the system corresponding to the reaction process of creating a Mahalanobis space of the reference at least one and, the on containing as data, abnormality existence determination method of measuring the reaction process, characterized by creating a Mahalanobis space as a reference index value of. In the method for determining the presence or absence of abnormality in the measurement reaction process according to claim 1,
A method for determining the presence / absence of an abnormality in a measurement reaction process, comprising the step of sequentially determining the presence / absence of an abnormality at each time point when the plurality of predetermined processes are completed. In the method for determining the presence or absence of abnormality in the measurement reaction process according to claim 1,
When calculating the Mahalanobis distance, information data (information data 1) for calculating the Mahalanobis distance and another new data (information data 2) orthogonal to the information data 1 are formed, and the two From the information data, a feature quantity capable of classifying the Mahalanobis distance into positive and negative is formed, and a classification or direction index is given to the Mahalanobis distance by the feature quantity. Presence judgment method. In the method for determining the presence or absence of abnormality in the measurement reaction process according to claim 1,
When calculating the Mahalanobis distance, information data (information data 1) for calculating the Mahalanobis distance and another new data (information data 2) orthogonal to the information data 1 are formed, and the two Forming a feature quantity that can represent the behavior or variation of each measurement item that is a constituent element of the Mahalanobis distance from one piece of information data, and performing the contribution or response magnification or ranking of each measurement item with the feature quantity, A method for determining the presence or absence of abnormality in a measurement reaction process, wherein pattern difference information of each item is given to the Mahalanobis distance. In the method for determining the presence or absence of abnormality in the measurement reaction process according to claim 1,
When calculating the Mahalanobis distance, information data (information data 1) for calculating the Mahalanobis distance and another new data (information data 2) orthogonal to the information data 1 are formed, and the two From the two pieces of information data, another new data (information data 2) orthogonal to the information data 1 is formed, and from the two pieces of information data, a feature quantity capable of classifying the Mahalanobis distance into positive and negative is formed, Providing an index of classification or directionality to the Mahalanobis distance by a feature amount;
A feature amount that can represent the behavior or variation of each measurement item that is a component of the Mahalanobis distance is formed, and the contribution amount, response magnification, or ranking of each measurement item is performed with the feature amount, and each Mahalanobis distance is Adding pattern difference information of items;
A method for determining the presence or absence of abnormality in a measurement reaction process, comprising: In the method for determining the presence or absence of abnormality in the measurement reaction process according to claim 1,
In the measurement, when the measurement result of a reference sample with a known concentration matches each concentration, and the operation / operation state of the mechanism unit or the system corresponding to the reaction process is normal or within the allowable use range The method for determining the presence or absence of abnormality in the measurement reaction process, characterized in that the reaction process data at that time is created as a normal data group. In the method for determining the presence or absence of abnormality in the measurement reaction process according to claim 5,
Abnormality in the measurement reaction process characterized by the Mahalanobis distance calculated from the data in the normal or allowable usage range of the operation / operation state of the mechanism unit or the system corresponding to the reaction process. Whether or not there is In the method for determining the presence or absence of abnormality in the measurement reaction process according to any one of claims 1 to 6,
A method for determining the presence or absence of an abnormality in a measurement reaction process, wherein identification information is added to measurement data in which an abnormality is detected when an abnormality in the reaction process is detected.   An automatic analyzer comprising storage means for storing a program for executing the method for determining the presence or absence of abnormality in a measurement reaction process according to any one of claims 1 to 6.   A storage medium storing a program for executing the abnormality determination method for a measurement reaction process according to any one of claims 1 to 6.

Images