Method for estimating cause of prolongation of blood clotting time, and information processing device

Abstract

The cause of the prolongation of the blood clotting time of the subject is estimated. A method for estimating the cause of the prolongation of blood clotting time acquires a clotting waveform (S301b) indicating a change in the amount of light over time caused by a clotting reaction of a reaction solution in which a test specimen and a reagent are mixed (S302b), and acquires a first waveform (S302b) by subjecting the clotting waveform to a pre-and post-differential treatment. First and second fitting waveforms (S303b, S304b) are obtained by fitting the solidification waveform and the first waveform, a third waveform (S305b) is obtained by front-back difference processing of the second fitting waveform, first to third normalized waveforms (S306b) are obtained by normalizing the light quantity axes and time axes of the first and second fitting waveforms and the third waveform, and the first to third normalized waveforms (S306b) are obtained by normalizing the light quantity axes and time axes of the first and second fitting waveforms and the third waveform. A feature amount is extracted from each of the first to third normalized waveforms (S307b), and the cause of the prolongation of the blood clotting time of the subject is estimated on the basis of the known feature amount and the extracted feature amount.

Description

Technical Field

[0001] The present disclosure relates to a method for estimating a cause of prolonged coagulation time and an information processing device. Background Art

[0002] Coagulation tests are performed for the purpose of understanding the condition of the coagulation and fibrinolytic system, diagnosing DIC (disseminated intravascular coagulation syndrome), confirming the effect of thrombotic treatment, and diagnosing hemophilia. In particular, the coagulation time measurement is to mix the specimen and the reagent and measure the time until the fibrin clot is formed (hereinafter referred to as the coagulation time). When there is an abnormality in the coagulation ability, either congenitally or acquired, the coagulation time will be prolonged. The causes of prolonged coagulation time include the lack of coagulation factors (deficiency type), the inhibition of coagulation reaction caused by antibodies against components in coagulation factors or reagents for coagulation time measurement (such as phospholipids) (inhibitor type), etc. Since the treatment policy is different for each cause of prolonged coagulation time, it is necessary to identify the cause of the prolongation, but the cause of the prolongation cannot be identified by simply measuring the coagulation time.

[0003] A representative method for identifying the cause of prolonged coagulation time is a cross-mixing test. In the cross-mixing test, the coagulation time of multiple plasma samples in which the test specimen and the normal specimen are mixed at different ratios (for example, 10:0, 9:1, 8:2, 5:5, 2:8, 1:9, 0:10) is measured. The measurement is performed immediately after the sample is prepared (immediate type) and after the sample is heated (cultured) for 2 hours (delayed type). The obtained coagulation time is plotted on the vertical axis, and the mixing ratio of the test specimen is plotted on the horizontal axis, and the graphs are connected with lines to form a graph, and the cause of the extension is determined based on the shape of the graphs of the immediate type and the delayed type. This judgment is qualitative, and skilled experience is required for the person making the judgment. In addition, the complexity of the preparation and culture operations of the specimen during the examination becomes a burden on the examiner. In addition, since additional blood sampling is required for this examination, there is also a burden on the patient.

[0004] As an evaluation method to reduce the burden on the examiner and the patient and provide more quantitative determination results, identification based on coagulation waveform analysis has been proposed. The coagulation waveform in question is a waveform obtained by optical measurement, and the turbidity change over time accompanying the formation of a fibrin clot is recorded.

[0005] Patent document 1 discloses a method for distinguishing whether the cause of the prolonged coagulation time in the test sample is the deficiency type or the inhibitor type by using a feature extracted from a coagulation waveform obtained from a plasma sample mixed with a test sample and a normal sample and its differential waveform. Patent document 2 discloses a method for distinguishing whether the presence of coagulation abnormality in the test sample and the activity value (concentration) of the coagulation factor are estimated by template matching using 50 parameters related to the center of gravity of the waveform obtained by first-order differentiation of the coagulation waveform obtained from the test sample. In particular, the method discloses a method for distinguishing which coagulation factor is estimated to be missing in the test sample by estimating the concentration of coagulation factor VIII (FVIII) and the concentration of coagulation factor IX (FIX). The qualitative and quantitative abnormalities of FVIII are called hemophilia A, and the qualitative and quantitative abnormalities of FIX are called hemophilia B. Which coagulation factor is missing is related to the choice of treatment policy, so it is important. In addition, Patent Document 3 discloses a system for estimating the main cause of APTT prolongation equipped with the above-mentioned template matching algorithm. In addition, in Non-Patent Document 1, the following method is disclosed: by using a flowchart of the slope, area, and time width ratio of a part of the waveform obtained by first-order differentiation of the coagulation waveform obtained from the test specimen, it is determined whether the test specimen is a deficiency type, an inhibitor type, or an anticoagulant is added. In particular, in the inhibitor type, it is determined whether the test specimen is a lupus anticoagulant (LA) positive group or an inhibitor expression group for FVIII. LA is one of the inhibitors for phospholipids, and its expression is related to antiphospholipid antibody syndrome. In addition, Patent Document 4 records: extracting multiple parameters from the first-order differential waveform and the second-order differential waveform of the coagulation waveform, and obtaining the expected concentration of each coagulation factor by multivariate correlation based on the multiple parameters; using a neural network learned for this prediction.

[0006] Prior Art Literature

[0007] Patent Literature

[0008] Patent Document 1: JP Patent No. 6994528

[0009] Patent Document 2: International Publication No. 2020 / 158948

[0010] Patent Document 3: JP Patent No. 6811975

[0011] Patent Document 4: U.S. Patent No. 6524861

[0012] Non-patent literature

[0013] Non-patent literature 1: D.Shimomura et al., The First-Derivative Curve of theCoagulation Waveform Reveals the Cause of aPTT Prolongation, Clinical andApplied Thrombosis / Hemostasis, Volume 26 (2020) P1-8 Summary of the invention

[0014] Problems to be solved by the invention

[0015] (1) The subject is to use characteristic quantities to identify whether the cause of the prolonged clotting time of a test specimen is a FVIII-deficient specimen group, a FIX-deficient specimen group, or a LA-positive specimen group, which are important in the diagnosis of hemophilia A, hemophilia B, or antiphospholipid antibody syndrome, wherein a coagulation reaction in a reaction solution generated by mixing the test specimen and a reagent is measured to obtain a coagulation waveform, and the obtained coagulation waveform is analyzed to extract the characteristic quantities.

[0016] (2) The more severe the disease, the longer the coagulation time of the specimen of the hemophiliac. In order to improve the throughput, the coagulation time measurement device appropriately determines that the coagulation reaction has ended for each specimen, thereby ending the measurement and moving on to the measurement of the next specimen. As a result, the data length of the coagulation waveform varies for each specimen.

[0017] Patent document 4 states that: based on multiple parameters, the expected concentration of each coagulation factor is obtained by multivariate posterior correlation; and a neural network is used in the estimation of the case, and the clustering method including the neural network uses the feature quantity of fixed data length in the input data. The currently known clustering methods such as neural network, mixed Gaussian distribution, K-means method, etc. are technologies based on the premise of fixed input data number. When the existing clustering methods including these are used in coagulation waveform analysis, the coagulation waveform data with variable data length is subjected to appropriate preprocessing and set to a fixed data length, so that the clustering method can function. As an example of such preprocessing for variable data length, data compression technology, zero padding technology, etc. are widely known as excellent preprocessing for image and sound recognition. If the cause of prolonged coagulation time can be estimated by a neural network using only the coagulation waveform of the test specimen or a series of coagulation waveform groups obtained in a cross-mixing test using a plurality of mixed specimens of the test specimen and a normal specimen at a given ratio, then the cause of prolonged coagulation time can be estimated by including not only the characteristic quantities determined experimentally but also the hidden information contained in the coagulation waveform. To achieve this, it is necessary to: perform preprocessing suitable for distinguishing coagulation waveforms as variable data length in a neural network; verify practical distinction accuracy by using a neural network using a coagulation waveform or a coagulation waveform group; and disclose specific values ​​of the structure and hyperparameters of the neural network used to achieve this.

[0018] Therefore, an object of the present disclosure is to estimate the cause of prolonged coagulation time of a sample using feature quantities extracted from a coagulation waveform indicating temporal changes in light quantity caused by a coagulation reaction of a reaction solution generated by mixing a sample and a reagent.

[0019] Furthermore, the present disclosure aims to estimate the cause of the extension of the blood coagulation time with practical accuracy using a neural network by performing appropriate preprocessing on a coagulation waveform having a variable data length.

[0020] Means for solving problems

[0021] The method for estimating the cause of prolonged coagulation time disclosed in the present invention comprises: obtaining a coagulation waveform representing the change in the amount of light over time caused by the coagulation reaction of a reaction solution, wherein the reaction solution is generated by mixing a test specimen composed of plasma obtained by separating blood obtained from a test subject, and a reagent; obtaining a first waveform by performing a front-to-back difference processing on the coagulation waveform; obtaining a first fitting waveform by performing a fitting processing on the coagulation waveform; obtaining a second fitting waveform by performing a front-to-back difference processing on the second fitting waveform; obtaining a second fitting waveform by performing a front-to-back difference processing on the second fitting waveform; obtaining a second fitting waveform by performing a front-to-back difference processing on the second fitting waveform; obtaining a second fitting waveform by performing a front-to-back difference processing on the second fitting waveform; obtaining a second fitting waveform by performing a front-to-back difference processing on the second fitting waveform; obtaining a first ... first fitting waveform by performing a front-to-back difference processing on the second fitting waveform; obtaining a first fitting waveform by performing a front-to-back difference processing on the second fitting waveform. 3 waveforms; obtaining a first normalized waveform, a second normalized waveform, and a third normalized waveform by normalizing the light quantity axis and the time axis of each of the first fitting waveform, the second fitting waveform, and the third waveform; extracting a feature quantity from each of the first normalized waveform, the second normalized waveform, and the third normalized waveform; and estimating the cause of the prolonged coagulation time of the test specimen based on the known feature quantity extracted from a specimen group whose cause of prolonged coagulation time is known and the feature quantity extracted from each of the first normalized waveform, the second normalized waveform, and the third normalized waveform.

[0022] In addition, the method for estimating the cause of prolonged coagulation time disclosed in the present invention estimates the cause of prolonged coagulation time of a test object based on a coagulation waveform representing the change in the amount of light over time caused by the coagulation reaction of a reaction solution, wherein the reaction solution is generated by mixing a test object composed of plasma obtained by separating blood obtained from a test object, and a reagent, and the estimation method comprises: making the data length of the coagulation waveform data consistent; using a neural network to estimate multiple causes of prolonged coagulation time from the data of the coagulation waveforms with consistent data lengths; and suggesting the cause of prolonged coagulation time of the test object based on the probability of belonging to the estimated cause of prolonged coagulation time.

[0023] Effects of the Invention

[0024] According to the present disclosure, the cause of prolonged coagulation time of a sample can be estimated using feature quantities extracted from a coagulation waveform indicating temporal changes in light quantity caused by a coagulation reaction of a reaction solution generated by mixing a sample and a reagent.

[0025] Furthermore, according to the present disclosure, by performing appropriate preprocessing on the coagulation waveform, which is data of variable length, the cause of prolonged coagulation time can be estimated with practical accuracy using a neural network.

[0026] Other problems, structures, and effects than those described above will become clear from the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a diagram showing the coagulation waveforms of two FVIII-deficient specimens in Example 1 and waveforms corresponding to their first-order differentials.

[0028] Figure 2 This is a diagram showing waveforms obtained by normalizing the coagulation waveforms, the waveforms corresponding to the first-order differentials of the coagulation waveforms, and the waveforms corresponding to the second-order differentials of the coagulation waveforms of the FVIII-deficient sample group, the FIX-deficient sample group, and the LA-positive sample group of Example 1.

[0029] Figure 3A This is a flowchart showing a method for estimating the cause of prolonged coagulation time in a comparative example.

[0030] Figure 3B This is a flowchart showing a method of estimating the cause of prolonged coagulation time in Example 1.

[0031] Figure 4 This is a diagram showing an overall configuration example of an automatic analyzer according to Example 1.

[0032] Figure 5 This is a diagram showing the waveforms WaveNor0, 1, 2 obtained by normalizing Wave0, 1, 2 in Example 1 and the feature vector.

[0033] Figure 6 This is a diagram showing a two-dimensional map composed of feature quantities of WaveNor0, 1, and 2 according to the first embodiment.

[0034] Figure 7 This is a diagram showing an example of plotting the characteristic amount of the sample to be tested on the two-dimensional map created based on the characteristic amount of the sample whose cause of prolongation of the coagulation time is known in Example 1.

[0035] Figure 8 It is a diagram showing normalized Waves 0, 1, and 2 of the test specimens of Example 1, a FVIII-deficient specimen group, a FIX-deficient specimen group, an LA-positive specimen group, a normal specimen group, an indicator showing an unknown belonging probability, and an example of a discrimination estimation result.

[0036] Fig. 9 This is a diagram showing an example of the relationship among the analysis unit, the PC, and the communication interface of the first embodiment, and the display of the results of the test sample on the display unit.

[0037] Fig.10 This is a diagram showing an example of WaveSrc1 and Wave1 in Example 1.

[0038] Fig.11A This is a diagram showing the neural network preprocessing method of Example 2.

[0039] Fig. 11B This is a diagram showing the neural network preprocessing method of Example 2.

[0040] Fig. 11CThis is a diagram showing the neural network preprocessing method of Example 2.

[0041] Fig.12 This is a diagram showing the structure of a neural network-based discriminator of Example 2.

[0042] Fig.13A The experimental results show the relationship between the pre-processing method of Example 2, the random dropout of the neural network-based discriminator, and the discrimination accuracy.

[0043] Fig. 13B The experimental results show the relationship between the pre-processing method of Example 2, the random dropout of the neural network-based discriminator, and the discrimination accuracy.

[0044] Fig. 13C The experimental results show the relationship between the pre-processing method of Example 2, the random dropout of the neural network-based discriminator, and the discrimination accuracy.

[0045] Fig.14A The experimental results show the relationship between the pre-processing method of Example 2, the hyperparameters of the neural network-based discriminator, and the discrimination accuracy.

[0046] Fig. 14B The experimental results show the relationship between the pre-processing method of Example 2, the hyperparameters of the neural network-based discriminator, and the discrimination accuracy.

[0047] Fig.15 It shows the experimental results of the pre-processing method of Example 2 and the neural network-based discriminator in suppressing over-learning.

[0048] Fig.16 These are experimental results showing the differences observed in the coagulation waveforms of Example 4 between the FVIII-deficient specimen and the FIX-deficient specimen.

[0049] Fig.17A This is an experimental result showing the relationship between the maximum value of the first-order differential, which is a well-known index, and the log coagulation time.

[0050] Fig. 17B It is the experimental result which shows the relationship between vec1_y and log coagulation time of Example 4.

[0051] Fig.18A This is the experimental result of Example 4 regarding WaveNor0 in which the coagulation waveforms of commercially available specimens were normalized based on the NRC method, and shows the representative waveforms of each specimen group.

[0052] Fig.18B WaveNor1 of Example 4 shows the experimental results of representative waveforms of each sample group, in which the coagulation waveforms of commercially available samples are normalized based on the NRC method.

[0053] Fig. 18C The experimental results of Example 4 show representative waveforms of each sample group with respect to WaveNor2 in which the coagulation waveforms of commercially available samples were normalized based on the NRC method.

[0054] Fig.19 This is the experimental result showing the frequency spectrum of WaveNor2 in Example 4.

[0055] Fig. 20 This is a diagram summarizing examples of feature quantity sets in the fourth embodiment.

[0056] Fig.21 This is the experimental result of Example 4 in which a group of commercially available samples is displayed on feature value space #6.

[0057] Fig.22A This is a display example of Example 4 showing the summary of commercially available sample groups plotted on the feature value space.

[0058] Fig. 22B This is an example of display of the identification results characterizing one FVIII-deficient specimen in Example 4.

[0059] Fig.23 This is a schematic diagram showing a method of reducing the dimension of a series of coagulation waveform data obtained for one specimen in a crossover test into two area indices F1 and F2 according to the fifth embodiment.

[0060] Fig.24 This is a diagram showing the results of sample identification in the cross-mixing test using multidimensional mixed Gaussian distribution in Example 5.

[0061] Fig.25A This is a diagram showing clustering areas of the indicators (F1, F2) of Example 5.

[0062] Fig.25B This is a diagram showing the experimental results of adding data points to the clustering areas of the indicators (F1, F2) of Example 5.

[0063] Fig.26 This is a diagram for explaining the indexes (G1, G2) of Example 5.

[0064] Fig.27A This is a graph showing experimental results of Example 5 in which the immediate (F1, G1) and delayed (F2, G2) values ​​of a combination of an index F calculated from a series of coagulation waveforms obtained from one specimen and an index G calculated from feature space #1 using the same specimen are plotted.

[0065] Fig.27B This is a diagram showing the experimental results of Example 5 in which the indices (F1, G1) and (F2, G2) obtained from 34 test samples are plotted.

[0066] Fig.28A 1 and 2 are experimental results showing the average shape of trajectory data for each cause of prolongation of the blood coagulation time obtained by the cross mixing test in Example 5.

[0067] Fig.28B This is a diagram showing the experimental results of Example 5 in which the trajectory data of each sample in the cross-mixing test was separated into a two-dimensional x14 space for identification.

[0068] Fig.29 This is a diagram summarizing the conditions for investigating the total number of specimens required for the identification method in Example 5.

[0069] Fig.30 This is a schematic diagram showing the relationship between the number of coagulation waveforms used in the identification of Example 5 and the worst probability (when the probability of the i-th specimen being classified into the correct specimen cluster is denoted as Pti, the minimum value of Pti among all specimens).

[0070] Fig.31 Example of display of identification results of the cross-mixing test of Example 5. DETAILED DESCRIPTION

[0071] Hereinafter, the present embodiment will be described with reference to the accompanying drawings. In the accompanying drawings, functionally identical elements are sometimes shown with the same number. In addition, the accompanying drawings show implementations and installation examples in accordance with the principles of the present disclosure, but these are used for understanding the present disclosure and are absolutely not used to interpret the present disclosure in a limiting manner. The descriptions in this specification are only typical examples and are not intended to limit the claims or application examples of the present disclosure in any sense.

[0072] In the present embodiment, the description is made in sufficient detail for those skilled in the art to implement the present disclosure, but it should be understood that other installations / modes are possible, and changes in structure / construction and replacement of various elements can be made without departing from the scope and spirit of the technical concept of the present disclosure. Therefore, the following description shall not be interpreted as being limited to this.

[0073] [Example 1]

[0074] In Example 1, the cause of prolonged coagulation time is estimated using feature quantities extracted from waveforms (WaveNor0, WaveNor1, WaveNor2) normalized based on the NRC (Normalization based on the Reaction Constants) method described later, which are obtained by waveform fitting.

[0075] The specimen used to obtain the coagulation waveform contains multiple coagulation factors. If the concentrations of the factors are different, there will be a difference in the speed of the coagulation reaction, which is related to the diversity (variability) of the coagulation time. In this field, the first-order differential of the coagulation waveform is treated as the coagulation speed, and the second-order differential is treated as the coagulation acceleration. Even in the case of the same specimen group, the coagulation waveform and the waveform related to the coagulation speed are not consistent, making it difficult to visually identify the characteristics of the specimen group. As an example, in Figure 1 The coagulation waveforms of two FVIII-deficient specimens (hereinafter, the coagulation waveforms are appropriately referred to as "WaveSrc0") and the first waveform corresponding to the first-order differential of the two coagulation waveforms (WaveSrc0) (hereinafter, the waveform corresponding to the first-order differential of the coagulation waveform is appropriately referred to as "WaveSrc1"). In WaveSrc0, the rising edge time of the waveform (arrow positions a1, a2), the total change in light intensity (arrow widths b1, b2), the coagulation time (for example, the light intensity ( Figure 1 The time c1, c2) when the change in scattered light intensity (counts) becomes 50% is different for each specimen. The total change in light volume depends on the concentration of fibrinogen, one of the coagulation factors. The coagulation time is affected by the speed of the coagulation reaction. In the case of hemophilia specimens, if the symptoms become severe, the coagulation time becomes longer, which is self-evident. In WaveSrc1, there are also differences in the height and width of the waveform for each specimen. If the coagulation time becomes longer, the waveform characteristics such as the coagulation speed (the change in the amount of light (scattered light intensity (counts)) per unit time) become smaller in inverse proportion to the coagulation time. This is also an experimental fact, which affects the difference in coagulation time and is different for each specimen. However, these two specimens are the same FVIII deficiency specimens, and common features must be found to distinguish the two specimens as a group of FVIII deficiency specimens.

[0076] The inventors believe that in order to make the common features in the same sample group obvious, it is effective to suppress the influence of the changes in the coagulation time within each sample group. Based on this idea, in the present disclosure, it is found that: by implementing normalization based on the NRC method that normalizes not only the light intensity axis but also the time axis at the same time, WaveSrc0 of the same sample group, WaveSrc1 obtained by the front-back difference processing of WaveSrc0, and WaveSrc2 obtained by the front-back difference processing of WaveSrc1 can converge, thereby making the waveform features within the same sample group obvious; when comparing the waveforms for each sample group, the difference in the waveform shape of each sample group can be made obvious. The specific method of normalization is described in detail later (Sample Identification Program). In addition, WaveSrc2 is a waveform equivalent to the first-order differential of WaveSrc1 (equivalent to the second-order differential of the coagulation waveform).

[0077] exist Figure 2Waveforms obtained by normalizing WaveSrc0, WaveSrc1, and WaveSrc2 of each of the FVIII-deficient sample group, the FIX-deficient sample group, and the lupus anticoagulant (LA)-positive sample group, namely WaveSrcNor0, WaveSrcNor1, and WaveSrcNor2, are shown. It can be seen that in any graph, the waveforms are roughly converged within the same sample group. In addition, if the same type of waveforms are compared with each other (for example, WaveSrcNor0 with each other) for each sample group, the shape difference can be seen. For example, in WaveSrcNor0, a difference can be seen in the rising edge position of the waveform in each sample group. The auxiliary line L1 is configured so that it is located at the center of the rising edge position of the waveform of the LA-positive group. The rising edge position of the waveform of the FVIII-deficient sample group is located to the right of the auxiliary line L1, and the rising edge of the waveform of the FIX-deficient sample group is located to the left of the auxiliary line L1, and the difference can be seen. In WaveSrcNor1 and WaveSrcNor2, the difference can be seen in the width between the vertices of the two peaks and the height of the vertices. In any waveform of WaveSrcNor1 and WaveSrcNor2, auxiliary lines L2 and L4 are configured at the center of the peak value of the peak on the left side of the LA positive group, and auxiliary lines L3 and L5 are configured to be located at the center of the peak value of the peak on the right side of the LA positive group. If the peak position of the peak on the right is confirmed with auxiliary lines L3 and L5 as a guide, no large difference can be seen in each sample group. On the other hand, if the peak position of the peak on the left is confirmed with auxiliary lines L2 and L4 as a guide, the peak value of the FVIII deficiency sample group is located at a position further to the right than the auxiliary lines L2 and L4, and the peak value of the FIX deficiency sample group is located at a position further to the left than the auxiliary lines L2 and L4, and the difference can be seen. Since the difference between these sample groups is greater than the variation within the same sample group, the inventors thought of using these differences as feature quantities and digitizing them for use in sample identification.

[0078] Figure 1 WaveSrc0 as shown and Figure 2The samples used in the acquisition of WaveSrc0, which is the source data for generating WaveSrcNor0 shown, are all commercially available products. Specifically, they are as follows. FVIII-deficient samples are Factor VIII Deficient Plasma (manufactured by Precision BioLogic, Inc.), FVIII Deficient Plasma (manufactured by George King Bio-Medical, Inc.), and Factor VIII Deficient Plasma (manufactured by Affinity Biologicals, Inc.). FIX-deficient samples are Factor IX Deficient Plasma (manufactured by Precision BioLogic, Inc.), FIX Deficient Plasma (manufactured by George King Bio-Medical, Inc.), and Factor IX Deficient Plasma (manufactured by Affinity Biologicals, Inc.). LA-positive samples are Lupus Positive Control and Weak Lupus Positive Control (manufactured by Precision BioLogic, Inc.), and Positive LA Plasma (manufactured by George King Bio-Medical, Inc.).

[0079] The specimen that is the object of the present disclosure is not particularly limited as long as it is a specimen that produces a coagulation reaction from the subject, and the specimen composed of plasma obtained by separating the blood obtained from the subject is suitable. In addition, the specimen composed of plasma can also contain trace impurities within the range that there is no problem for the analysis in the automatic analyzer of the present disclosure, such as the measurement of the coagulation waveform. In addition, the reagent is not particularly limited as long as it is a reagent for measuring the thromboplastin time (PT) project, the activated partial thromboplastin time (APTT) project, and the fibrinogen (Fbg) project. In addition, the device can be a structure that can measure the change of the light amount (the so-called light amount is an indicator including turbidity) of the reaction solution caused by the coagulation reaction over time. The reagent used in the measurement of the coagulation waveform shown in this embodiment is Coagpia APTT-N (manufactured by Sekisui Medical Technology Co., Ltd.), and the device is Hitachi Automatic Analyzer 3500 (manufactured by Hitachi High-Tech Co., Ltd.).

[0080] The present disclosure is characterized not only in that normalization based on the NRC method is performed to make the difference in waveform shape more obvious, but also in that waveform fitting is performed to numerically convert the difference in waveform shape that has become obvious to suppress the influence of noise.

[0081] Next, an automatic analyzer for acquiring a coagulation waveform, a method for acquiring a coagulation waveform, a sample identification program, and the like will be described in detail.

[0082] (Automatic Analyzer 100)

[0083] The automatic analyzer 100 used for acquiring the coagulation waveform will be described. Figure 4 1 is a diagram showing an example of the overall configuration of the automatic analyzer 100 according to the first embodiment. Figure 4 The basic device operation is described below, but is not limited to the following examples.

[0084] like Figure 4 As shown, the automatic analyzer 100 is roughly composed of an analysis unit 130, an operation computer 118, a storage unit 119, and a control computer 120. The analysis unit 130 is roughly composed of a sample dispensing mechanism 101, a sample disk 102, a reagent dispensing mechanism 106, a reagent disk 107, a reaction container storage unit 111, a reaction container transport mechanism 112, a detection unit 113, and a reaction container disposal unit 117. The control computer 120 is connected to the analysis unit 130 so as to be communicable.

[0085] The specimen dispensing mechanism 101 aspirates specimens 103a contained in specimen containers 103 arranged on a specimen disk 102 rotating clockwise and counterclockwise, and discharges them into reaction containers 104. The specimen dispensing mechanism 101 aspirates and discharges specimens by the operation of a specimen syringe pump 105 controlled by a control computer 120.

[0086] The reagent dispensing mechanism 106 sucks the reagent 108a contained in the reagent container 108 arranged on the reagent disk 107 and discharges it into the reaction container 104. The reagent dispensing mechanism 106 performs the reagent suction operation and discharge operation by the operation of the reagent syringe pump 110 controlled by the control computer 120.

[0087] Furthermore, a reagent temperature increasing mechanism 109 is built in the reagent dispensing mechanism 106 , and the temperature of the reagent 108 a sucked by the reagent dispensing mechanism 106 is increased to an appropriate temperature (predetermined temperature) by the reagent temperature increasing mechanism 109 controlled by the control computer 120 .

[0088] The reaction container transport mechanism 112 transports and sets the reaction container 104. The reaction container transport mechanism 112 transports and sets the reaction container 104 from the reaction container storage unit 111 to the reaction container setting unit 114 of the detection module 113 by holding the reaction container 104 and rotating it in the horizontal direction.

[0089] The detection unit 113 has a reaction container setting part 114 for placing one or more reaction containers 104 (in this embodiment, one case is shown as an example), and measures the light intensity of the reaction liquid (mixed liquid of the specimen 103a and the reagent 108a) in the reaction container 104 inserted into the reaction container setting part 114. The detection unit 113 is temperature-controlled so that the reaction liquid in the reaction container 104 inserted into the reaction container setting part 114 is, for example, 37°C. In addition, in this embodiment, a case where one detection unit 113 is configured is shown, but it can also be configured to have a plurality of detection units 113. An example of the detection principle in the detection unit 113 is described below. The light irradiated from the light source 115 is scattered by the reaction liquid in the reaction container 104. The detection unit (light sensor) 116 receives the scattered light scattered by the reaction liquid in the reaction container 104. For example, a halogen lamp or an LED is used as the light source 115. The detection unit (light sensor) 116 is composed of a photodiode or the like. The signal received by the detection unit (optical sensor) 116 becomes a light quantity value converted into a digital signal by the A / D converter 121, and is input to the control computer 120 as reaction process data (coagulation waveform data representing the time change of the detection light quantity), and is taken into the storage unit 119. The operation of the detection unit 113 is controlled by the control computer 120. Here, a detector using the scattering of light is used, but there are also detectors using transmitted light.

[0090] The reaction container transport mechanism 112 holds the reaction container 104 after measurement and discards it to the reaction container discarding unit 117 .

[0091] For the purpose of improving the throughput, an incubator 122 without a detector may be provided for heating the sample before adding the measurement start reagent.

[0092] The control computer 120 is an information processing device having a computer system having a processor, a memory, etc., and controls the operation of the automatic analyzer such as the dispensing of the sample 103a and the reagent 108a, the transfer of the reaction container 104, and the disposal of the reaction container 104, and calculates the coagulation time based on the measured value of the light intensity (coagulation waveform data) that changes with time corresponding to the coagulation reaction of the reaction solution, and also performs identification and estimation of the cause of the extension of the coagulation time. The calculated coagulation time and the identification and estimation results of the cause of the extension of the coagulation time are output to the display unit 118c and stored in the storage unit 119. In addition, these results can be printed out by the printer 123 via the operation computer 118. In the storage unit 119 connected to the control computer 120, in addition to the control program, the measurement program, the calibration curve generation program, the quantitative program, the sample identification program, etc., the measured waveform data, the quantitative results, the sample identification results, etc. are stored. Various programs are read out and executed according to the request input to the operating computer 118 or the request sent from the communication interface 124. The input to the operating computer 118 can be input by touching the display unit 118c, or can be input via the connected keyboard 118b. In addition, the content displayed on the display unit 118c can be selected by using the mouse 118a for input. The waveform data, quantitative results, specimen identification results, etc. stored in the storage unit 119 are output to the display unit 118c. Alternatively, they are sent to the communication interface 124. Alternatively, they are printed using the printer 123 as needed. The communication interface 124 is connected to the network within the hospital, for example, and communicates with HIS (Hospital Information System) and LIS (Laboratory Information System).

[0093] As described above, the sample identification program and the waveform data of the sample stored in the storage unit 119 are called by the control computer 120 to perform estimation of the cause of the prolonged coagulation time of the sample. Alternatively, the waveform data of the sample may be forwarded to the analysis computer 125 that executes the sample identification program, and the analysis computer 125 may estimate the cause of the prolonged coagulation time of the sample, and the result may be displayed on the screen of the analysis computer 125. Alternatively, the waveform data stored in the storage unit 119 may be written to other external storage media via the control computer 120 and the operation computer 118, and an independent analysis computer having the sample identification program may read the written waveform data to estimate the cause of the prolonged coagulation time of the sample.

[0094] (How to obtain coagulation waveform)

[0095] Next, the analysis operation of the coagulation time item and the acquisition of the coagulation waveform (WaveSrc0) are explained. The coagulation time items are mainly the thromboplastin time (PT) item, the activated partial thromboplastin time (APTT) item, and the fibrinogen (Fbg) item. In any item, the parameters required for the analysis are set first. In the parameter setting, the item to be analyzed, the sample amount, the reagent amount, the output unit, etc. are input. In the PT item and the Fbg item, the calibration method is also input, and calibration is performed as needed. After setting the parameters, the sample container 103 and the reagent container 108 that respectively contain the sample 103a and the reagent 108a are respectively set to the sample disk 102 and the reagent disk 107 for analysis. The reagent 108a can use commercially available PT measurement reagents, APTT measurement reagents, and Fbg measurement reagents, for example.

[0096] (Analysis action of PT project)

[0097] The analysis action of the PT item is described. The sample 103a is dispensed into the empty reaction container 104 contained in the reaction container storage section 111 by the sample dispensing mechanism 101. The reaction container 104 containing the sample 103a is moved to the reaction container setting section 114 of the detection component 113 by the reaction container transport mechanism 112. The reagent 108a is sucked from the reagent container 108 by the reagent dispensing mechanism 106 and heated to a suitable temperature in the reagent heating mechanism 109. The temperature at this time is expected to be 37°C. The heated reagent 108a is ejected into the reaction container 104 containing the sample 103a that has been set in the reaction container setting section 114. At the same time as the ejection starts, the turbidity change over time of the mixed solution of the sample 103a and the reagent 108a, that is, the reaction solution, is optically measured to become WaveSrc0.

[0098] (Analysis of APTT project)

[0099] The analysis action of the APTT project is described. The sample 103a is dispensed into the empty reaction container 104 contained in the reaction container storage part 111 by the sample dispensing mechanism 101. The reaction container 104 containing the sample 103a is moved to the reaction container setting part 114 of the detection component 113 by the reaction container transport mechanism 112. The first reagent contains an activator such as ellagic acid and kaolin, is sucked from the reagent container 108 by the reagent dispensing mechanism 106, and is heated to a suitable temperature in the reagent heating mechanism 109. The desired temperature at this time is 37°C. The heated first reagent is ejected into the reaction container 104 containing the sample 103a which has been set in the reaction container setting part 114. The mixed solution of the sample 103a and the first reagent is temperature-controlled in the reaction container 104. The desired temperature at this time is 37°C. At this time, the mixed solution of the sample 103a and the first reagent can also be stirred by a stirring mechanism not shown. The second reagent is a calcium chloride solution, which is sucked from the reagent container 108 by the reagent dispensing mechanism 106 and heated to a suitable temperature by the reagent heating mechanism 109. The temperature at this time is expected to be 37°C. The heated second reagent is ejected into the reaction container 104 containing the mixed solution of the specimen 103a and the first reagent, which is set in the reaction container setting part 114. The mixed solution of the specimen 103a and the first reagent is temperature-controlled for a certain time until the ejection of the second reagent, and the time is expected to be 180 seconds. At the same time as the ejection of the second reagent begins, the turbidity change over time of the mixed solution of the specimen 103a and the first and second reagents, that is, the reaction solution, is optically measured to become WaveSrc0.

[0100] (Analysis action of Fbg project)

[0101] The analysis action of the Fbg item is described. The specimen 103a is dispensed into the empty reaction container 104 contained in the reaction container storage section 111 by the specimen dispensing mechanism 101. The reaction container 104 containing the specimen 103a is moved to the reaction container setting section 114 of the detection component 113 by the reaction container transport mechanism 112. The specimen diluent is sucked from the reagent container 108 by the reagent dispensing mechanism 106 and heated to a suitable temperature by the reagent heating mechanism 109. The desired temperature at this time is 37°C. The heated specimen diluent is ejected into the reaction container 104 containing the specimen 103a which has been set in the reaction container setting section 114. The mixed solution of the specimen 103a and the specimen diluent is temperature-controlled in the reaction container 104. The desired temperature at this time is 37°C. At this time, the mixed solution of the specimen 103a and the specimen diluent may also be stirred by a stirring mechanism (not shown). The reagent is sucked from the reagent container 108 by the reagent dispensing mechanism 106 and heated to a suitable temperature in the reagent heating mechanism 109. The desired temperature at this time is 37°C. The heated reagent is ejected into the reaction container 104 containing the mixed solution of the specimen 103a and the specimen diluent which is set in the reaction container setting part 114. At the same time as the ejection of the reagent starts, the turbidity change over time of the mixed solution of the specimen 103a, the specimen diluent and the reagent, i.e., the reaction solution, is optically measured to become WaveSrc0.

[0102] Regardless of the item, the measurement is terminated at the time point when the coagulation reaction is completed (when the turbidity change is no longer observed). Alternatively, the measurement is terminated after a certain time has passed. The so-called certain time is, for example, 5 minutes. The turbidity change over time measured by light is taken into the storage unit 119 as a coagulation waveform via the control computer 120.

[0103] An example of a coagulation waveform Figure 1 As shown, the horizontal axis represents the measurement time, and the vertical axis represents the light intensity value. Figure 1 This is an example of a coagulation waveform of a FVIII-deficient specimen (Precision BioLogic Inc.) in the APTT test. The vertical axis represents the scattered light intensity.

[0104] (Specimen identification procedure)

[0105] Figure 3B 1 is a flowchart showing a method for estimating the cause of prolonged coagulation time by executing the sample identification program. The computer system of the control computer 120 performs the estimation of the cause of prolonged coagulation time of Example 1 by executing the sample identification program. Figure 3B To explain in detail. Figure 3BEach step of the flowchart may be executed by the control computer 120, the operation computer 118, the analysis computer 125, or another computer on the network. Here, an example in which the computer system of the control computer 120 executes each step is described.

[0106] The control computer 120 reads the coagulation waveform data (WaveSrc0) obtained by measuring the coagulation reaction into ( Figure 3B - S301b). The specific analysis operation related to the acquisition of WaveSrc0 is as described in the above (Method of Acquiring Coagulation Waveform).

[0107] Next, the control computer 120 generates WaveSrc1 by performing a front-to-back difference process on WaveSrc0 ( Figure 3B - S302b). WaveSrc1 is obtained, for example, by Formula 1. Here, Δt in the following Formula 1 is a measurement interval of the coagulation reaction, for example, 0.1 seconds.

[0108] [Formula 1]

[0109]

[0110] Next, the control computer 120 generates a first fitting waveform Wave0 (hereinafter, a waveform obtained by fitting the coagulation waveform is referred to as the first fitting waveform and is appropriately referred to as “Wave0”) by waveform fitting of WaveSrc0. Figure 3B - S303b), Wave1 is generated by waveform fitting of WaveSrc1 (hereinafter, the waveform obtained by fitting the first waveform is referred to as the second fitting waveform and is appropriately referred to as "Wave1") Figure 3B -S304b). The fitting formula for WaveSrc0 is based on the fact that the coagulation waveform is a Sigmoid curve (S-shaped curve), and it is desirable to use a formula based on the Sigmoid curve. For example, it is Formula 2. In Formula 2, after the reagent is ejected, the light amount becomes Y dc The formula of the coagulation waveform.

[0111] [Formula 2]

[0112]

[0113] Y in the above equation 2 dc is the basic light quantity [counts], Y0 is the basic scattered light quantity [counts] of the reaction solution itself, Y1 is the saturated scattered light quantity [counts], t is the measurement time [seconds], T0 is the time of (Y0+Y1)÷2 [seconds], and T1 is the reaction time constant [seconds].

[0114] Formula 2 assumes the fitting of the coagulation waveform measured by scattered light. However, in the case of the coagulation waveform measured by absorbance, the waveform measured by scattered light is inverted upside down. Therefore, Formula 2 can be improved so as to follow the coagulation waveform measured by absorbance.

[0115] The fitting formula for WaveSrc1 is expected to be a formula that can better reproduce the bimodal property. For example, the peak on the left side of the bimodal property can be represented by Gaussian distribution, the peak on the right side can be represented by Sigmoid differential, and the two can be connected by a third-order spline function. The shape of WaveSrc1 is often bimodal, and the tendency to become this shape is particularly significant in specimens with prolonged coagulation time.

[0116] Next, the control computer 120 generates Wave 2 by performing a front-to-back differential process on Wave 1 (hereinafter, a waveform corresponding to the first-order differential of the first fitting waveform is referred to as a third waveform and is appropriately referred to as “Wave 2”) ( Figure 3B -S305b). Wave2 is obtained by, for example, Formula 3.

[0117] [Formula 3]

[0118]

[0119] Next, the control computer 120 normalizes both the time axis and the light intensity axis of Wave 0, 1, and 2 based on the NRC method ( Figure 3B -S306b). The time axis is normalized by Formula 4. As the light intensity axis, Wave0, 1, and 2 are normalized by Formula 5, Formula 6, and Formula 7, respectively. The waveform obtained by normalizing the time axis and the light intensity axis of Wave0 is the first normalized waveform, which is appropriately referred to as "WaveNor0" hereinafter. In addition, the waveform obtained by normalizing the time axis and the light intensity axis of Wave1 is the second normalized waveform, which is appropriately referred to as "WaveNor1" hereinafter. In addition, the waveform obtained by normalizing the time axis and the light intensity axis of Wave2 is the third normalized waveform, which is appropriately referred to as "WaveNor2" hereinafter. Normalization is based on reaction constants, called NRC (Normalization based on the Reaction Constants) method. The reaction constant used in the NRC method can use the numerical value obtained by fitting WaveSrc0. In order to suppress the influence of noise, the difference is digitized and extracted as a unique feature value, and waveform fitting is implemented.

[0120] [Formula 4]

[0121]

[0122] [Formula 5]

[0123]

[0124] [Formula 6]

[0125]

[0126] [Formula 7]

[0127]

[0128] Here, n is the number of data points (n ≧ 1), and the relationship between t(n) and n is t(n) = n / 10 [seconds]. In addition, T0, T1, Wave0 (n), Y dc , Y1 is obtained from formula 2.

[0129] Finally, the control computer 120 extracts the feature quantity ( Figure 3B -S307b). The feature quantity must be a product of numerically quantifying the part where the waveform shape difference can be seen according to the sample group. Figure 2 .exist Figure 2 The waveform (WaveSrcNor0, 1, 2) obtained by normalizing the original waveform (WaveSrc0, 1, 2) before waveform fitting is shown in FIG. Normalization follows equations 4 to 7, but Figure 2 The original waveform is normalized, and Wave0(n) in Formula 5, Wave1(n) in Formula 6, and Wave2(n) in Formula 7 are substituted into WaveSrc0, WaveSrc1, and WaveSrc2, respectively.

[0130] As mentioned above, for example, in WaveSrcNor0, the rising edge position of the waveform can be seen to be different. The order of the rising edge is the order of the FIX-deficient specimen group, the LA-positive specimen group, and the FVIII-deficient specimen group from the side with the smallest normalized time. If the formula 4 of the normalized formula of time is followed, the smaller the reaction time constant T1 (i.e., the faster the reaction), the farther the distance from the normalized time 0. That is, the difference in the reaction speed of each specimen group is characterized as the difference in the timing of the rising edge. FVIII becomes activated FVIII, and because its presence enhances the reaction of a part of the coagulation reaction mechanism, it is believed that in the FVIII-deficient specimen group, the reaction time constant becomes larger and the rising edge timing slows down. LA has the property of hindering the coupling of coagulation factors due to coupling with the scaffold of the coagulation reaction, i.e., phospholipids, and therefore, uniformly hinders the coagulation reaction regardless of the activity of the coagulation factor. Since the degree of obstruction varies according to the amount and quality (titer) of LA, the speed of the coagulation reaction becomes diverse, which is related to the diversity of the reaction time constant T1, and is considered to be between the FIX-deficient sample group and the FVIII-deficient sample group. In this way, it is considered that the rising edge timing reflects the difference due to the extension of the coagulation time, and it is judged that it is useful to quantify the difference in the rising edge timing. Here, a vector with a point where the normalized time of each waveform is 0 as the starting point and a rising point of the waveform as the end point is set as a feature quantity.

[0131] In addition, in WaveSrcNor1, the difference can be seen in the width between the vertices of the two peaks and the height of the vertices. In the FVIII-deficient specimen group, because the FVIII present in a trace amount is used at the beginning of the reaction, it is insufficient afterwards, therefore, it is believed that the first peak becomes larger (higher). On the other hand, in the FIX-deficient specimen group and the LA-positive specimen group, because FVIII fully exists, therefore, compared with the initial stage of the reaction, in the amplification reaction called positive feedback produced thereafter, the reaction speed becomes faster, and it is believed that the second peak becomes larger (higher). In addition, if focusing on the normalized time, there is a difference in the timing of the rising edge of the waveform for the same reason as WaveSrcNor0, which is related to the difference in the normalized time of the first peak. It can be considered that the difference in the generation time of the first peak and the difference in the height of the two peaks reflect the difference due to the extension of the cause of the clotting time, and it is judged that the characteristic amount of quantifying these differences is useful. Here, a vector having a point at which the normalized time is 0 as a starting point and a peak point of the first peak as an end point is set as a feature amount.

[0132] In WaveSrcNor2, the difference between the first peak height and its generation time can be seen in the form of being affected by the waveform difference equivalent to the first-order differential. Here, the peak height of the first peak and the normalized time information are combined, and the vector information starting from the point where the normalized time is 0 and ending at the peak point of the first peak is set as the feature amount.

[0133] In order to digitize the data while suppressing the influence of noise, vector features are extracted from the waveform (WaveNor0, 1, 2) obtained by normalizing the fitting waveform (Wave0, 1, 2). Figure 5 The waveforms obtained by normalizing Wave0, 1, and 2, namely WaveNor0, 1, 2, and the characteristic vector are shown. The horizontal axis represents the normalized time, and the vertical axis represents the normalized level. The characteristic vectors of WaveNor0, 1, and 2 are set to vectors 0, 1, and 2, respectively. The x-component and y-component of each vector are set to the characteristic values ​​(X, Y) of each waveform. Here, the x-component of WaveNor0 is worked on so that T0 / T1 is subtracted to emphasize the difference in the waveform shape of the specimen group. The characteristic values ​​(X, Y) of WaveNor0, 1, and 2 are respectively expressed as (vec0_x, vec0_y), (vec1_x, vec1_y), and (vec2_x, vec2_y).

[0134] Finally, the extracted feature values ​​of the sample to be tested are compared with the feature values ​​of the sample whose sample background (cause of prolonged coagulation time) is known to identify the cause of prolonged coagulation time of the sample to be tested. The feature values ​​(X, Y) extracted from the sample whose cause of prolonged coagulation time is known are used independently, and in the two-dimensional map in which these feature values ​​are plotted, the feature value groups (clusters) of each sample group have a probability density distribution. For example, a probability density distribution based on a mixed Gaussian distribution model is made. In Figure 6 An example of a two-dimensional mapping consisting of feature quantities extracted from normalized waveforms, namely WaveNor0, 1, and 2, is shown. In the mapping, it is expressed as Wave0, 1, and 2. The feature quantity is a feature quantity extracted by analyzing the waveform obtained from a specimen whose cause of prolongation of the coagulation time is known. The ellipses illustrated for the groups (clusters) of feature quantities of each specimen group (FVIII deficiency, FIX deficiency, LA positive, normal) represent the extension of the feature quantity data. The size of each ellipse is set to 4σ (σ: standard deviation in each direction, called Mahalanobis distance in two dimensions or more) from the center so that the ellipses of normal specimens and abnormal specimens do not overlap. These ellipses are represented by the probability density of Gaussian distribution. In the figure, the area that does not belong to any ellipse indicates that no specimen group belongs to it within the range of the given data, corresponding to the situation of artificially mixing more than two kinds of specimens to be tested, the status of drug administration of patients under treatment, etc., indicating that it is a new coagulation waveform. This area is defined as an "unknown" group. The ellipses of abnormal specimens overlap in any discriminator, and the boundary can be set to an arbitrary value of the belonging probability calculated according to the probability density distribution. exist Figure 6In the figure, the case where the probability of belonging to the boundary is 50% is shown. The abnormal specimens used here are the commercial products mentioned above. Normal specimens are also replaced by commercial products. Specifically, they are Normal Reference Plasma and Pooled Normal Plasma (manufactured by Precision BioLogic, Inc.), Factor Assay Control Plasma and Borderline Factor Assay Control Plasma (manufactured by George King Bio-Medical, Inc.).

[0135] The feature quantities extracted from WaveNor0, 1, and 2 of the sample under test are applied to these three two-dimensional maps to calculate the probability of belonging to the sample under test in each map (the probability of belonging to which sample group). The sample under test is identified and estimated according to the calculated belonging probability. For example, the sample group estimated to have the highest belonging probability can be identified, the sample group estimated to have the highest belonging probability by averaging the belonging probabilities of each mapping can be identified, and the sample group estimated to have the highest belonging probability by weighted average of the belonging probabilities of each mapping can be identified.

[0136] exist Figure 7 as well as Figure 8 An example of a display showing the identification and estimation results of the test specimen is shown. Figure 7 This is a diagram showing an example of where the feature quantity of the sample to be tested is plotted in a two-dimensional map created based on the feature quantity of a known sample. Figure 8 It is a graph showing WaveNor0, 1, 2 (labeled as Wave0, Wave1×2, Wave2 in the figure) of the test specimen, the FVIII-deficient specimen group, the FIX-deficient specimen group, the LA-positive specimen group, the normal specimen group, an indicator showing the probability of unknown belonging, and the discrimination estimation result. Fig. 9 An example of the relationship among the analysis unit 130, the control computer 120 and the communication interface 124, and the display of the test sample results on the display unit 118c is shown. The display unit 118c displays the sample number (S_No.) of the sample for which the results are desired to be displayed, and a column for two-dimensional mapping. It is preferable that the display unit 118c displays the coagulation time, the result of the identification estimation, and the estimated result of the coagulation factor activity of the selected sample number. The estimated result of the coagulation factor activity does not necessarily have to be displayed, but it is better to display it. For example, the so-called coagulation factor VIII deficiency is a state in which the activity value of the coagulation factor VIII is less than 1%, and the so-called coagulation factor IX deficiency is a state in which the activity value of the coagulation factor IX is less than 1%. Fig. 9In the figure, the selected map and the characteristic quantity of the sample in the map are plotted on the lower left side of the screen. By displaying it in this way, the cause of the prolonged coagulation time of the sample can be intuitively expressed with high visual recognition. On the other hand, it is also important to quantitatively express the cause of the prolonged time. Fig. 9 In the example, “Hemophilia A (probability: 95%)” is shown as “prediction result”. Fig. 9 As shown in the lower right side of the screen, the waveform and the details of the probability of belonging in the sample of the selected sample number are displayed (for example Figure 8 It is better to make the waveform visible and show an indicator showing the details of the probability to which it belongs.

[0137] In Example 1, for example, in the reading of WaveSrc0 ( Figure 3B -S301b), and / or after the generation of WaveSrc1 ( Figure 3B -S302b) and then appropriately perform pre-processing such as smoothing. Pre-processing (the first filtering process to remove the noise of WaveSrc0, and the second filtering process to remove the noise of WaveSrc1) means that each waveform is filtered for the purpose of reducing noise. This processing step does not necessarily have to be performed. However, in order to perform waveform fitting with better accuracy, it is desirable to implement it before waveform fitting. As a specific example of filtering, there are smoothing processes such as N-point moving average processing (N is a positive integer and an odd number, 3≦N≦the number of measurement points), and selection of the minimum value in a certain interval. In addition, efforts can be made to make the T1 point moving average of the reaction time constant using the fitting parameters based on Formula 2 without losing the original waveform shape relative to the change in the value of the coagulation time. At this time, for example, the decimal point of the value of T1 is rounded off to an integer, and in the case of an even number, 1 is added or subtracted to use it as an odd number.

[0138] (Example of the execution result of the sample identification program)

[0139] The results of identifying the presumed test specimens in the specimen identification program can distinguish the FVIII-deficient specimen group, the FIX-deficient specimen group, and the LA-positive specimen group with 85.7% (=12 / 14*100%), 100% (=9 / 9*100%), and 66.7% (=4 / 6*100%), respectively. As a whole, the probability is 86% (=25 / 29*100%). Normal specimens can be identified 100% (=10 / 10*100%), and it is confirmed that normal specimens will not be misjudged as FVIII-deficient specimen group, FIX-deficient specimen group, and LA-positive specimen group. The test specimens used are products with batch differences of the above-mentioned commercial products.

[0140] (Effects of Example 1)

[0141] exist Fig.10 The example of WaveSrc1 and Wave1 is shown. The waveform of WaveSrc1 can be seen to contain noise. As a method of removing noise, pre-processing such as smoothing is disclosed, but in excessive processing, there is a problem related to the loss of waveform shape, and in insufficient processing, there is a problem that when a certain feature is extracted as a numerical value, there are multiple values ​​and it cannot be uniquely determined. When the inherent feature of the coagulation waveform is extracted as a numerical value, there are multiple values ​​and it cannot be uniquely determined, for example, when it is desired to extract Fig.10 In the case of the maximum value of the left peak of the two peaks, in the raw data containing noise, the maximum value exists at multiple points at different times and cannot be determined at one point. The method of extracting feature quantities by waveform fitting determines waveform model parameters from the observed waveform containing noise so that, for example, the RMS error becomes the minimum, and then numerically calculates the inherent features in the coagulation waveform from the fitting waveform without noise, and solves such a problem through such a process.

[0142] In addition, the cause of the prolonged coagulation time is estimated using the extracted feature quantity. Specifically, the feature quantity is extracted using a specimen whose cause of the prolonged coagulation time is known, and in a two-dimensional map in which the feature quantity is plotted, a feature quantity group (cluster) for each specimen group is given a probability density distribution. The feature quantity of the specimen to be tested is applied to the two-dimensional map to calculate the probability of belonging to the specimen (the probability of belonging to which specimen group), and the cause of the prolonged coagulation time of the specimen to be tested can be intuitively estimated with high visual recognition according to the belonging probability.

[0143] In the present disclosure, a new method for solving the above-mentioned problems of the prior art is proposed in the following aspects: the intrinsic characteristics of the coagulation waveform are extracted as numerical values ​​through the medium of a fitting model; and the characteristics of each cause of prolongation of the coagulation time are made clear by normalizing it using the NRC method based on the reaction constant.

[0144] Here, in Figure 3A An example of the sample identification process involved in the comparative example is shown in FIG. Figure 3A ), firstly, the coagulation waveform data obtained by coagulation reaction measurement is read into ( Figure 3A -S301a). Next, pre-processing such as smoothing is performed ( Figure 3A -S302a). Next, a normalization process (sometimes also called a correction process) is performed. Figure 3A -S303a). Normalization is performed only on the light intensity axis. Next, the time differential of the waveform data (sometimes only differential) or further time differential is calculated to calculate the waveform related to the coagulation rate or coagulation acceleration ( Figure 3A-S304a). Finally, the waveform parameters are extracted as feature quantities ( Figure 3A -S305a), for example, the identification and estimation of the specimen to be tested is performed based on the correlation between the characteristic quantity of the specimen whose cause of the extension of the coagulation time is known and the characteristic quantity of the specimen to be tested. At this time, the pre-processing and the normalization of the light intensity axis do not necessarily have to be performed, and the process can be skipped. In addition, although not shown in the figure in the process, sometimes the axis representing the light intensity change rate of the waveform related to the coagulation rate or coagulation acceleration is normalized, and the case of implementing normalization with the maximum value of each waveform set to 100% is disclosed.

[0145] In contrast, the present disclosure is significantly different from the comparative example in that waveform fitting is performed, both the time axis and the light intensity axis are normalized, and intuitive identification estimation can be performed from the extracted feature quantity with high visual recognition. Figure 3B ), first, the coagulation waveform (WaveSrc0) data obtained by coagulation reaction measurement is read into ( Figure 3B -S301b). Next, WaveSrc1 is generated by performing the front-to-back difference processing of WaveSrc0 ( Figure 3B -S302b). Next, Wave0 is generated by waveform fitting of WaveSrc0 ( Figure 3B -S303b), generate Wave1 by waveform fitting of WaveSrc1 ( Figure 3B -S304b). Next, Wave2 is generated by performing the difference processing of Wave1 before and after ( Figure 3B -S305b). WaveSrc1 or Wave1 in the present disclosure corresponds to the waveform related to the solidification speed in the comparative example, and Wave2 corresponds to the waveform related to the solidification acceleration in the comparative example. Next, Wave0, 1, and 2 are normalized based on the NRC method ( Figure 3B -S306b). The so-called normalization based on the NRC method in the present disclosure is normalization based on the reaction constant, which refers to normalization of both the time axis and the light intensity axis. As disclosed in the comparative example, when normalizing the light intensity axis, the waveforms related to the coagulation waveform, the coagulation rate, or the coagulation acceleration are not normalized with the maximum value of each waveform. In the present disclosure, normalization is performed to make Wave0, 1, and 2 correlated. Finally, the feature quantity ( Figure 3B - S307b), using the feature quantity of the sample to be tested in a two-dimensional map having a probability density distribution created using feature quantities extracted from a sample whose cause of prolonged coagulation time is known, the identification estimation of the sample to be tested is performed. In the present disclosure, for example, in the reading of WaveSrc0 ( Figure 3B -S301b) and / or after the generation of WaveSrc1 ( Figure 3B- S302b) After that, pre-processing such as smoothing is appropriately performed.

[0146] The main creativity of the present disclosure lies in that the normalization in the time axis direction is performed by utilizing the reaction constant measured by the fitting model using WaveSrc0, and the time differential of the coagulation waveform disclosed in Patent Documents 1 to 4 and Non-Patent Document 1 (recorded only as differential) is replaced by the difference of the time constant contained in the reaction constant (hereinafter referred to as the reaction time constant difference), thereby making it possible to estimate the cause of the prolonged coagulation time while reducing the influence of the severity of the symptoms manifested by the length of the coagulation time.

[0147] [Example 2]

[0148] In Example 2, identification of a coagulation waveform using a neural network is described (from the perspective of the commonality of the invention, this is limited to an example using WaveSrc0).

[0149] Here, an embodiment of a method for estimating the cause of prolonged coagulation time by a neural network using coagulation waveform data as input is presented, based on the time axis normalization method disclosed in the present invention. As mentioned above, the estimation of the cause of prolonged coagulation time by a neural network directly using coagulation waveform data has not been disclosed as a set of techniques that can be easily reproduced by a person skilled in the art, and therefore, the disclosure of the following three points becomes the gist of this embodiment.

[0150] (1) Since the measured coagulation waveform is variable length data, a pre-processing technique suitable for a neural network is disclosed.

[0151] (2) To demonstrate that the clustering of coagulation waveforms performed by the neural network is practical and to confirm that the discrimination accuracy is >90%.

[0152] (3) In order to easily reproduce the above-mentioned neural network and facilitate its expansion to the application of measurement data from various devices and the application of samples from patients undergoing treatment, structural parameters such as the number of nodes and the number of hidden layers, as well as hyperparameters such as the number of epochs and the number of batches involved in the learning conditions are provided.

[0153] (1) Disclosure of pre-processing

[0154] First, the pre-processing technology suitable for neural networks is described.

[0155] Figures 11A to 11C The graph shows the experimental results of the coagulation waveform (WaveSrc0) of a commercially available specimen and the coagulation waveform preprocessed to be suitable for a neural network. Fig.11AIt is WaveSrc0 of a commercially available specimen. Here, 103 waveforms consisting of a FVIII-deficient specimen group (46 waveforms), a FIX-deficient specimen group (22 waveforms), an LA-positive specimen group (6 waveforms), and a normal specimen group (29 waveforms) were used. The measuring device used in the determination was a Hitachi automatic analyzer 3500 (manufactured by Hitachi High-Technologies Co., Ltd.). The vertical axis becomes the count value of the quantized data of the A / D converter corresponding to the scattered light intensity.

[0156] Fig. 11B The waveform of WaveSrc0 is represented as a pre-processing of a neural network by a zero-filling method known in the field of image and sound recognition. Here, according to the main inventiveness of the present disclosure, that is, the normalization of the time axis using the reaction time constant and the response time, the time axis is normalized, the range of -5τ to +10τ is set to 501 points of equally spaced data, and the coagulation waveform is resampled by linear interpolation. On the vertical axis, the value at the beginning of the measurement is zero, and normalization is performed so that the change in scattered light intensity of 10,000 counts is set to 1. This normalization is used to adapt to general Sigmoid and hyperbolic tangent as the activation function of the neural network. The normalization used here is used to use the change in scattered light intensity as a feature value, and the normalization is the formula 8 that is different from the above-mentioned formula 5.

[0157] [Formula 8]

[0158]

[0159] In addition, in the time range of -5τ to +10τ, the data "0" is used to supplement the places where there is no data due to the interruption of measurement, so as to make the data length consistent.

[0160] Fig. 11C The waveform data represents the case where the fitting model is used to supplement the insufficient data. If the parameters τ, T0, Y0, and Y1 obtained by the fitting model of Formula 2 are used, it is easy to obtain continuous numerical data even for the moment after the measurement is completed. In this sense, the introduction of the fitting model can be said to have a high affinity for the discrimination of the coagulation waveform using the neural network. As a simple method following this, the value of the measured final data point can also be copied to supplement the insufficient data. Fig. 11C and Fig. 11B In comparison, the data representing the scattered light intensity after the measurement end point is smoothed. As a preprocessing used in the neural network, the data length becomes fixed. Fig. 11B , Fig. 11C It is suitable.

[0161] (2) The structure of the neural network and the demonstration of its practical performance

[0162] Next, the structure of the neural network is described.

[0163] Fig.12 It shows the structure of the neural network of the present disclosure. Fig.12 It is a generalization of Keras contained in TensorFlow, which is common in the research on neural networks. As can be seen in the figure, the neural network disclosed in the present invention can provide the practical performance shown previously through a structure with 4 hidden layers, 512 nodes, and tight coupling of all layers. The number of learning rounds is 10,000, the batch size is 32, and the hyperbolic tangent is used as the activation function. Regarding the activation function, using Sigmoid instead of hyperbolic tangent can also obtain the same discrimination performance. This should be easily understood by technicians involved in this field.

[0164] In the following, as hyperparameters, the number of rounds = 10000, the batch size = 32 and fixed are used for explanation. In addition, the coagulation waveform data of commercially available specimens is divided into 60% for learning, 20% for verification, and 20% for testing, and the cross entropy is used as the loss function. The reason for dividing the data into three parts is explained. Since the parameters such as weights that constitute the neural network are automatically updated relative to the learning data, the hyperparameters such as the number of hidden layers, the number of nodes, random inactivation, the number of rounds, and the batch size are appropriately selected to minimize the cross entropy of the verification data. In order to objectively evaluate the performance of the neural network determined by these, it is necessary to evaluate the cross entropy and discrimination accuracy using test data that was not used in the previous steps.

[0165] Figures 13A to 13C This is an experimental result showing the relationship between the learning result and random dropout. Here, the result with the number of nodes = 256 and the number of hidden layers = 3 is shown. Fig.13A Is used through Fig. 11B The results of the coagulation waveforms in which the data lengths are made uniform by the zero-padding method are shown. It can be seen that even if the random dropout amount that is effective in suppressing overlearning is changed, the discrimination accuracy of the test data does not exceed 90%.

[0166] Fig. 13B Is used through Fig. 11C The results of the case where the coagulation waveform was pre-processed by data extrapolation by fitting model are shown. As can be seen from the figure, under the condition that the random inactivation amount is 0.4 or less, the practical discrimination accuracy target of 90% or more can be obtained. Therefore, it can be proved that the pre-processing method disclosed in the present invention and the specimen discrimination method based on the coagulation waveform analysis using the structure of the neural network are discrimination methods that can be easily reproduced and widely used by technicians in this field.

[0167] Fig. 13CThis is the result of using a coagulation waveform whose data length is made consistent by complementation based on the replication of the final data point of the measurement. As before, it can be seen that the discrimination accuracy is 90% or more under the condition that the random inactivation amount is 0.4 or less. Fig. 13B Compared with the interpolation method shown, it has the advantage of less computational effort. When the sample data is further increased or noise is mixed into the measurement for some reason, it can be said that data extrapolation by fitting model is excellent in the stability of discrimination performance. In the following, unless otherwise specified, preprocessing is based on data extrapolation by fitting model.

[0168] (3) Structural parameters such as the number of nodes and hidden layers, as well as hyperparameters such as the number of rounds and batches involved in the learning conditions

[0169] Next, ranges of parameters suitable for implementation of the present disclosure are described.

[0170] The number of rounds suitable for learning a neural network, random dropout, etc. are as described above. Here, the number of nodes and examples are described as numerical values ​​representing the structure of a neural network.

[0171] Fig.14A as well as Fig. 14B These are experimental results showing the range of hyperparameters in which a discriminator combining the preprocessing method of the present disclosure and a neural network shows practical performance. Fig.14A The relationship between the number of hidden layers and the discrimination accuracy is shown. Here, the number of nodes is set to 1024 and the random dropout is set to 0.25. As can be seen from the figure, in the discrimination method disclosed in the present invention, when the number of hidden layers is 3 or more, a practical discrimination accuracy of more than 90% can be obtained.

[0172] Fig. 14B This is an experimental result showing the relationship between the number of nodes and the discrimination accuracy. Here, the case where the number of hidden layers is 3 is shown. As can be seen from the figure, in the discrimination method disclosed in the present invention, when the number of nodes is 16 or more, a practical discrimination accuracy of 90% or more can be obtained. Based on these results, it is shown that the discrimination method disclosed in the present invention can obtain a practical discrimination accuracy by setting the number of hidden layers to 3 or more and the number of nodes to 16 or more.

[0173] Fig.15These are experimental results showing that the pre-processing method and the neural network-based discrimination method disclosed in the present invention achieve sufficient suppression of over-learning. Here, the relationship between the number of nodes and the cross entropy for learning, verification, and test samples is shown. In the field of voice and face recognition, which is widely used as a neural network, the cross entropy is about 0.1 to 0.3. In contrast, it can be seen that in the discrimination method disclosed in the present invention, when the number of nodes is more than 16, the cross entropy becomes about 0.1 or less, and the cross entropy value is roughly the same for each sample group for learning, verification, and testing. This is a result showing that over-learning can be fully suppressed by the structure of the neural network shown in the present invention.

[0174] As described above, according to the present disclosure, after the coagulation waveform obtained from the specimen is pre-processed using the reaction time constant and response time quantified by the fitting model so as to have a given data length, a neural network having 3 or more hidden layers and 16 or more nodes can be used to estimate the cause of the extension of the coagulation time with practical discrimination accuracy. In addition, it goes without saying that since the neural network-based discriminator outputs the probability of belonging to each extension cause cluster of the coagulation time as an output, the neural network-based discriminator shown here can also cope with Figure 8 The display of the probability of belonging to the example is self-evident.

[0175] Furthermore, by applying the NRC method to the waveform data used for learning, it is possible to further improve the discrimination accuracy.

[0176] (Effects of Example 2)

[0177] By performing appropriate preprocessing on the coagulation waveform, which is data of variable length, the cause of prolonged coagulation time can be estimated with practical accuracy using a neural network.

[0178] [Example 3]

[0179] In Example 3, the use of a cross-mixing test is described.

[0180] The above describes the identification of the cause of prolonged coagulation time using the coagulation waveform obtained by measuring the specimen collected from the subject. On the other hand, in the case where the coagulation time of the specimen collected from the subject is prolonged, a cross-mixing test is recognized as a representative method for identifying the cause of prolonged coagulation time, wherein in the cross-mixing test, the coagulation time of a plurality of mixed plasma samples in which the specimen and a normal specimen are mixed at a given ratio is measured, and the cause of prolonged coagulation time is screened by plotting the shape of a graph of the coagulation time according to the mixing ratio of the specimen. According to the present disclosure, based on a series of coagulation waveforms obtained from a plurality of mixed plasma samples, not only the mixing ratio dependence of the coagulation time can be obtained, but also a plurality of characteristic quantities caused by the cause of prolonged coagulation time contained in the series of coagulation waveforms and the probability of belonging to each prolonged cause can be obtained. Therefore, the specimen identification method of the present disclosure can be easily applied to the coagulation waveform data obtained in the cross-mixing test, and it is expected to seek to improve the identification accuracy more than before. When the emphasis is placed on compatibility with existing methods, for example, the following method can be said to be preferred: quantifying the mixing ratio dependence of the coagulation time, and simultaneously quantifying the coagulation factor VIII deficiency and the coagulation factor IX deficiency based on the coagulation waveform of the specimen collected from the subject, and mapping them into a two-dimensional space.

[0181] [Example 4]

[0182] In Example 4, the feature quantity of the coagulation waveform obtained by normalization processing using the NRC method based on the reaction constant is expanded.

[0183] That is, in Example 4, feature quantities are extracted not only from the first normalized waveform (WaveNor0), the second normalized waveform (WaveNor1), and the third normalized waveform (WaveNor2), but also from waveform data obtained by performing nonlinear operations on these waveforms, and the cause of the prolongation of the coagulation time of the test specimen is estimated based on the distribution information of the feature quantities of each cause of prolongation of the coagulation time prepared in advance and the extracted feature quantities.

[0184] The method for estimating the cause of the prolonged clotting time of the test subject in Example 4 includes: (A) obtaining a clotting waveform representing the change in the amount of light over time caused by the clotting reaction of the reaction solution, where the reaction solution is generated by mixing the test subject composed of plasma separated from the blood obtained from the subject and a reagent; (B) obtaining a first waveform by performing differential processing on the clotting waveform; (C) obtaining a first fitting waveform by performing fitting processing on the clotting waveform; (D) obtaining a second fitting waveform by performing fitting processing on the first waveform; (E) obtaining a third waveform by performing forward and backward differential processing on the second fitting waveform; (F) obtaining a first normalized waveform, a second normalized waveform, and a third normalized waveform by normalizing the light amount axis and the time axis of the first fitting waveform, the second fitting waveform, and the third waveform respectively; (G) extracting a feature quantity from at least one of the first normalized waveform, the second normalized waveform, the third normalized waveform, and the waveform data obtained by performing a non-linear operation on these; (H) estimating the cause of the prolonged clotting time of the test subject based on the distribution information of the feature quantities for each cause of the prolonged clotting time prepared in advance and the feature quantity extracted from the clotting waveform of the test subject.

[0185] Hereinafter, in the present invention, for the sake of simplifying the figures and helping with the understanding of the technology, the FVIII-deficient specimen group is respectively given an attached reference numeral and called "FVIII deficiency", the FIX-deficient specimen group is given an attached reference numeral and called "FIX deficiency", the LA-positive specimen group is given an attached reference numeral and called "LA", and the normal specimen group is given an attached reference numeral and called "Normal". Here, a clotting waveform measured by a Hitachi automatic analyzer 3500 (manufactured by Hitachi High-Technologies Corporation) using commercially available specimens is used. The measurement reagent is Coagpia APTT-N (manufactured by积水医疗科技株式会社). In addition to the commercially available products described in Example 1, Coagpia control P-NI (manufactured by积水医疗科技公司) and Coagtrol N (manufactured by希森美康公司) are used as normal specimens.

[0186] Fig.16 It is a graph comparing WaveSrc0 of one representative specimen each of the FVIII-deficient specimen and the FIX-deficient specimen. As Fig.16 can be seen, regarding the clotting time, both are approximately 100 seconds and are roughly the same. On the other hand, it can be seen that the convergence of the reaction of the FVIII-deficient specimen is slower compared to the FIX-deficient specimen (the time from the initial value to the saturation value of the scattered light intensity is longer). This characterizes the difference in the influence of the lacking factors on the reaction process for both. This feature is, for example, that T0 for both is approximately equal in Equation 2, and T1 is FIX deficiency < FVIII deficiency. In the present disclosure, (T0 / T1) is subtracted from vec0_x.

[0187] In addition, when the initial value of the scattered light intensity is set to 0% and the saturation value is set to 100%, the coagulation time is calculated as the time until the level reaches 50%, but the calculation method of the coagulation time is not limited to this method. For example, the time until reaching an arbitrary level other than 50% may be calculated as the coagulation time.

[0188] Fig.17A This is an experimental result showing the relationship between the maximum value of the first-order differential of the coagulation waveform and the log coagulation time. The maximum value of the first-order differential is equivalent to the index |min1| described in Patent Document 1. Fig.17A As shown, the R2 value of the least squares fit with the second-order function is 0.9323. That is, it can be seen that the maximum value of the first-order differential is highly correlated with the coagulation time. It can be seen that the maximum values ​​of the first-order differentials of the FVIII-deficient sample group, the FIX-deficient sample group, and the LA-positive sample group (LA) in the figure are distributed in a very small area compared with the overall change amount of the index including the normal sample group.

[0189] Fig. 17B This is an experimental result showing the relationship between vec1_y and log coagulation time in Example 4. The R2 value of the least squares fit with a second-order function is 0.12. That is, it can be seen that there is almost no correlation between vec1_y and coagulation time in Example 4. The distribution of vec1_y in the FVIII-deficient specimen group, the FIX-deficient specimen group, and the LA-positive specimen group (LA) in the figure is a larger change than the overall change of the index including the normal specimen group. In addition, it can be seen that each specimen group is separated in an independent region where the value of vec1_y becomes the relationship of FVIII deficiency>LA>FIX deficiency.

[0190] The results are also similar to those of the maximum value of the second-order differential of the coagulation waveform obtained by scattered light measurement, which corresponds to another index |min2| described in Patent Document 1. Fig.17A Likewise, as is clear from the above experimental facts, it is known that by normalization based on the NRC method of the present disclosure, a waveform independent of the coagulation time is generated, and by extracting feature quantities, the performance of distinguishing the cause of prolonged coagulation time is improved compared with the existing indexes.

[0191] Fig.18A The experimental results of the representative waveforms of each sample group are shown in WaveNor0, which is a coagulation waveform of a commercially available sample normalized based on the NRC method in Example 4. It is found that the difference between the sample groups is large in the rising edge of the waveform.

[0192] Fig.18BThe experimental results of the representative waveforms of each sample group are shown in WaveNor1, which is the coagulation waveform of the commercially available sample normalized based on the NRC method in Example 4. Similarly, it can be seen that the difference in the waveform of each sample group is clearly represented.

[0193] Fig. 18C The experimental results of WaveNor2 showing the representative waveforms of each sample group are shown for the coagulation waveforms of commercially available samples normalized based on the NRC method in Example 4. Similarly, it can be seen that the differences in the waveforms of each sample group are clearly characterized.

[0194] Fig.19 This is an experimental result showing the spectrum of WaveNor2 of Example 4. Here, for the representative specimen of each specimen group, i=0 to 2047 WaveNor2 waveform arrays are generated in the normalized time interval -50T1 to +50T1, and Fourier transformation is performed. As the feature quantity of each specimen group, for example, the amplitude spectrum value of frequency FREQ1 (minimum value of HB, i=40) and FREQ2 (maximum value of HA, i=115) can be added.

[0195] Fig. 20 This is a diagram summarizing examples of feature quantity sets of Example 4. Generally, if there are more feature quantities, it becomes possible to identify the causes of prolonged coagulation time from different viewpoints, so it is preferred. However, there are limitations here. For example, when there are feature quantities α and β, it is not preferred to add γ=aα+bβ (a and b are constants) to the feature quantities. This is because γ is a linear coupling of α and β. Even if γ is added to the feature quantity, the amount of information used for identification will not increase, and the identification of the causes of prolonged coagulation time from different viewpoints cannot be implemented. Furthermore, due to the addition of γ, the influence of measurement noise and numerical calculation errors increases, thereby damaging the reliability of identification of the causes of prolonged coagulation time. On the other hand, the waveform data obtained by mapping the normalized waveform (WaveNor0, WaveNor1, WaveNor2) to other spaces (converted into new information) through nonlinear operations such as exponentiation, differentiation, Fourier transformation, Galois transformation, and convolution operations is simple to increase the effective feature quantity because the mathematical orthogonality relative to the original data is guaranteed. Of course, it is also possible to select feature quantities by listing multiple feature quantity candidates from the same waveform data, obtaining the mutual correlation coefficients of all their combinations, and excluding feature quantities with large mutual correlation coefficients (the amount of additional information for identification becomes less). The following is an example of adding values ​​obtained from waveform data mapped to other independent spaces. As shown in the figure, here, two quantities can be used as feature quantities from each of the seven feature quantity spaces (#0 to #6).

[0196] (1) Feature space #0

[0197] These are vec0_x-(T0 / T1) and vec0_y in Example 1.

[0198] (2) Feature space #1

[0199] These are vec1_x and vec1_y in Example 1.

[0200] (3) Feature space #2

[0201] These are vec2_x and vec2_y in Example 1.

[0202] (4) Feature space #3

[0203] It is the RMS error between WaveNor2 and the representative waveform of the FIX-deficient sample group, and the RMS error between WaveNor2 and the representative waveform of the normal sample group.

[0204] (5) Feature space #4

[0205] These are the x- and y-coordinates of the intersection of WaveNor1 and WaveNor2.

[0206] (6) Feature space #5

[0207] yes Fig.19 The amplitude spectrum values ​​of FREQ1 and the amplitude spectrum values ​​of FREQ2 in the frequency spectrum of WaveNor2 are shown.

[0208] (7) Feature space #6

[0209] When t50 represents the coagulation time, it is a numerical value (Severity_Index) defined by the empirical formula representing the severity of the disease as shown in the following equation 9, and the RMS error between WaveNor1 and the representative waveform of the normal sample group.

[0210] The representative waveform may be an average waveform, which is obtained by normalizing waveforms obtained from a group of specimens having the same level of activity of coagulation factors based on the NRC method and averaging normalized levels at each normalization time.

[0211] [Formula 9]

[0212]

[0213] In addition, Fig. 20 In order to facilitate the implementation of numerical analysis, the recommended x and y ranges when illustrating each feature space are attached. In the following, in Example 4, the diagram is disclosed according to these ranges.

[0214] In Example 4, as a feature quantity identification process, in each feature quantity space, identification is performed in accordance with the probability density distribution based on the two-dimensional mixed Gaussian distribution. As for the basic mixed Gaussian distribution, it is well known to those skilled in the art, so it is not explained here. In Example 4, in order to improve the reliability of the identification result, the mixed Gaussian distribution that can physically internalize the influence of the measurement error according to the central limit theorem is used as the basis, and the upper limit of the Mahalanobis distance is given to the probability density distribution of each cluster. The Mahalanobis distance is a quantity equivalent to the standard deviation of the one-dimensional Gaussian distribution. If it is called in units of σ according to convention, the upper limit of 3 to 6σ is appropriate. By assigning the upper limit of the Mahalanobis distance, the basic shape of each cluster area on the feature quantity space becomes an ellipse. When two clusters are adjacent, the boundary is a condition where the probability of belonging to each cluster becomes 50%. The data of one specimen is displayed as one point in each feature quantity space, and the probability of belonging to the cluster of each specimen group is calculated at the same time. The identification process of Example 4 is implemented by averaging the probability of belonging to the clusters in each feature space, and the cluster of the specimen group to which the specimen belongs with the highest probability and the probability of belonging are obtained as the identification result. In this embodiment, an example of obtaining the probability of belonging to each specimen group by averaging the probability of belonging to the specimen group in each feature space is shown, but for example, the probability of belonging to the cluster in each feature space can be weighted averaged to identify the specimen group with the highest probability of belonging, or the identification can be performed based on the probability of belonging to the specimen group in any one of the feature spaces.

[0215] The upper limit of the so-called Mahalanobis distance introduced in Example 4 means that, for example, a data point that is more than 3σ away from the center of a sample group cluster does not belong to the cluster when comparing the standard deviation of the measured data. In identification by general statistical distribution, SVM, K-means, deep neural network, etc., since the space is divided into a specified number of clusters, new or abnormal data is classified into any known cluster.

[0216] Regarding the data generation of each cluster, the distribution data consisting of the center position and height and the variance-covariance matrix can be calculated based on 2 feature value data each extracted from the coagulation waveform data of the FVIII-deficient specimen (FVIII-deficient), the FIX-deficient specimen (FIX-deficient), the LA-positive specimen (LA), and the normal specimen (Normal) in the 7 feature value spaces, thereby forming a cluster for each specimen group.

[0217] An additional note will be made regarding feature space #6. Fig.21 This is the experimental result of displaying the commercially available sample group of Example 4 on feature space #6. Fig.21As shown, it uses Severity_Index shown in Formula 9 as the x-axis feature value, and the other feature value space identifies the clustering of 4 sample groups, while it is dedicated to the identification of normal samples and other than this. Among the commercially available samples prepared, Borderline Factor Assay Control Plasma (GeorgeKing Bio-Medical, Inc.) with a coagulation time of 42 seconds and LA-positive samples with a coagulation time of 46 seconds were separated by a Mahalanobis distance of more than 3σ, and it was confirmed that they could be identified almost perfectly. The FVIII activity value and FIX activity value of Borderline Factor Assay Control Plasma are 41% and 55% respectively based on the attached analytical values. This sample cannot be classified into any of the LA-positive sample group, the coagulation factor inhibitor-positive sample group, the FVIII-deficient sample group, and the FIX-deficient sample group. Therefore, in this embodiment, for convenience, it is set to the range of normal samples for treatment. By changing the upper limit of the Mahalanobis distance, the normal range can be adjusted.

[0218] Fig.22A This is a display example showing the summary of the commercially available sample group of Example 4 plotted on the feature value space. Fig.22A In the figure, the commercially available sample group used in the experiment of Example 4 is plotted on the seven feature quantity spaces and the WaveNor1 obtained by normalization and the correct answer rate of the identification results are summarized. Here, the upper limit of the Mahalanobis distance is set to 4σ. The correct answer rate is 100% (106 / 106*100%).

[0219] Fig. 22B This is an example of displaying the identification results of a FVIII-deficient specimen in Example 4. In each feature value space of Example 4, the specimen is represented as one point, and the probability of belonging to the cluster of each specimen group is shown. In this case, it is FVIII-deficient (83%), LA (17%), FIX-deficient (0%), and Normal (0%).

[0220] [Example 5]

[0221] In Example 5, a specific method for applying the technology of the present invention in a cross-mixing test is described in accordance with the outline shown in Example 3. Feature quantities are extracted from the first normalized waveform (WaveNor0), the second normalized waveform (WaveNor1), and the third normalized waveform (WaveNor2) of a series of sample specimens of plasma and normal plasma separated from blood obtained from a subject at a plurality of mixing ratios, and feature quantities are extracted from waveform data obtained by performing nonlinear operations on these, and the cause of the prolonged coagulation time of the subject is estimated using at least one of these feature quantities.

[0222] Regarding the cross-mixing test, as described in paragraph 0121 of Patent Document 1, "The changes in the graph patterns of LA-positive specimens and coagulation factor inhibitor-positive specimens are similar. Therefore, it is known that the discrimination between the two based on the changes in the graph patterns has to be a qualitative evaluation, and it is difficult to determine if one is not an expert." It is a well-known fact that it is difficult to discriminate between LA-positive specimens and coagulation factor inhibitor-positive specimens based on the relationship between the mixing ratio of normal plasma and test plasma and the coagulation time. In this embodiment, the subject is to provide the following method: applying the solution technology of subject (1) to the cross-mixing test, not only can the LA-positive specimen group and the coagulation factor inhibitor-positive specimen group be discriminated, but also the FVIII-deficient specimen group and the FIX-deficient specimen group be included, so that even if one is not an expert, the cause of the prolonged coagulation time can be quantitatively discriminated.

[0223] The method for estimating the cause of prolonged coagulation time of a test subject of Example 5 comprises: (A) preparing a series of sample specimens obtained by mixing plasma and normal plasma separated from blood obtained from a test subject at multiple mixing ratios; (B) obtaining a series of coagulation waveforms representing the change in light intensity over time caused by the coagulation reaction of a reaction solution generated by mixing the series of sample specimens and a reagent; (C) obtaining a series of first waveforms by performing front-to-back difference processing on the series of coagulation waveforms; (D) obtaining a series of first fitting waveforms by performing fitting processing on the series of coagulation waveforms; (E) obtaining a series of second fitting waveforms by performing fitting processing on the series of first waveforms; and (F) obtaining a series of second fitting waveforms by performing fitting processing on the series of first waveforms. (G) obtaining a series of first normalized waveforms, a series of second normalized waveforms, and a series of third normalized waveforms by performing front-to-back differential processing of a series of second fitting waveforms; (H) extracting a feature value from at least one of a series of first normalized waveforms, a series of second normalized waveforms, a series of third normalized waveforms, and a series of waveform data obtained by performing nonlinear operations on these waveforms; and (I) using at least one of a series of feature values ​​extracted from a series of coagulation waveforms to infer the cause of the extension of the coagulation time of the sample specimen.

[0224] In addition, the computer system such as the control computer 120 described in the first embodiment executes each of the processes (A) to (I) described above.

[0225] Hereinafter, in this embodiment, as a little help for simplifying the diagram and understanding the technology, the FVIII-deficient specimen group is referred to as "FVIII-deficient", the FIX-deficient specimen group is referred to as "FIX-deficient", the LA-positive specimen group is referred to as "LA", and the coagulation factor inhibitor-positive specimen group (here, a commercially available product with coagulation factor inhibitors added to the FVIII-deficient specimen group) is referred to as "FVIII-deficient inh". Commercially available specimens were used in the experiment. Specifically, as follows. The FVIII-deficient specimen is Factor VIII Deficient Plasma (manufactured by Precision BioLogic, Inc.). The FIX-deficient specimen is Factor IX Deficient Plasma (manufactured by Precision BioLogic, Inc.). The LA-positive specimens are Lupus Positive Control and Weak Lupus Positive Control (manufactured by Precision BioLogic, Inc.). The coagulation factor inhibitor positive specimens are Mild Factor VIII Inhibitor Plasma and Strong Factor VIII Inhibitor Plasma (manufactured by Affinity Biologicals, Inc.). The normal specimen is Pooled Normal Plasma (manufactured by Precision BioLogic, Inc.). Here, FVIII deficiency specimens, FIX deficiency specimens, LA positive specimens, and coagulation factor inhibitor positive specimens are used as test specimens, respectively. In this embodiment, by treating the data of repeated measurements as one specimen, a total of 34 test specimens are considered (FVIII deficiency: 7, FIX deficiency: 7, LA: 8, FVIII deficiency inh: 12). The measuring device is Hitachi automatic analyzer 3500 (manufactured by Hitachi High-Technologies Co., Ltd.), and the reagent is CoagpiaAPTT-N (manufactured by Sekisui Medical Technology Co., Ltd.).

[0226] In this embodiment, a series of plasma samples of a test specimen and a normal specimen mixed at different ratios (10:0, 9:1, 8:2, 5:5, 2:8, 1:9, 0:10) are prepared, and the coagulation waveform is measured immediately after the sample is prepared (immediate type), and the coagulation waveform is measured after the sample is warmed at 37°C for 2 hours (delayed type). For one test specimen, a series of 14 coagulation waveforms (7 series of immediate type and delayed type) are measured, and the feature quantity extracted from the waveform normalized based on the NRC method is calculated. In the following, in order to characterize the mixed plasma sample, 7 mixing ratios (0.0, 0.1, 0.2, 0.5, 0.8, 0.9, 1.0) and measurement conditions (immediate or delayed) of the test specimen are used.

[0227] Fig.23 t1 is a schematic diagram showing a method of reducing the dimension of a series of coagulation time data obtained for one test specimen in a cross-mixing test into two area indices F1 and F2 in Example 5. In a cross-mixing test, there is a condition where the coagulation time changes little with respect to the change in the mixing ratio of the test specimen. In this case, the obtained coagulation time data is strongly affected by the mixing conditions and measurement errors and is biased. Therefore, the average value of the two coagulation times (immediate and delayed) in the case of mixing ratio = 0.0 is set to t2. s , the average of the two solidification times (immediate and delayed) for the case of mixing ratio = 1.0 is set as t e , the difference from the linear approximation coagulation time is numerically converted into area indices F1 and F2 based on the secondary formula, thereby implementing dimension reduction and indexing by reducing the deviation of the measurement results based on averaging. In the secondary formula, n is an integer (n=1, 2), N is the number of the mixing ratio (here N=7), i is an exponential integer representing the mixing ratio, and t i is the solidification time, t^ i According to t s and t e The linear approximation of the solidification time, Δx i is the mixing rate x with the i-th i The corresponding integration interval is to be i-1 、x i+1 The width of each midpoint (x in the case of the left end) i-1 = 0.0, and the right side is set to x i+1 = 1.0). Under the experimental conditions implemented in this embodiment, relative to the mixing rate x i = (0.0, 0.1, 0.2, 0.5, 0.8, 0.9, 1.0) and becomes Δx i =(0.05, 0.1, 0.2, 0.3, 0.2, 0.1, 0.05).

[0228] [Formula 10]

[0229]

[0230] If the index F1 of the immediate measurement and the index F2 of the delayed measurement are calculated according to the above formula, for example, even if the mixing ratios used in the immediate measurement and the delayed measurement are different, the same dimensional indexes can be obtained by setting -1≦F1≦+1 and -1≦F2≦+1.

[0231] Fig.24 This is a diagram showing the results of specimen identification in a cross-mixing test using a multidimensional mixed Gaussian distribution in Example 5. Here, a case is described in which a series of 14 test specimens are modulated to change the mixing ratio of one source specimen with a normal specimen, and the immediate and delayed time conditions, and then tested. The method disclosed in Example 1 is a case in which one coagulation waveform is associated with one test specimen that is a source. If it is expanded to a case in which 14 coagulation waveforms are associated with one source test specimen, 28 feature quantities are quantified in each feature quantity space. In each feature quantity space, the two-dimensional mixed Gaussian distribution shown in Example 1 is expanded to a 28-dimensional mixed Gaussian distribution to quantify the identification accuracy. In addition, as a conventional example, the original data of the coagulation time is identified with a 14-dimensional mixed Gaussian distribution, and the dimensionality reduction indicators (F1, F2) shown previously are identified with a two-dimensional mixed Gaussian distribution, and the results are summarized. As Fig.24 As can be seen in the figure, the discrimination accuracy is in the range of 32.4% to 88.2%. The discrimination accuracy is the lowest when the raw data of the coagulation time is used, and the discrimination accuracy is the highest when the index (F1, F2) is used. The two are essentially the same, but as mentioned above, since (F1, F2) reduces the influence of the deviation, a higher discrimination accuracy can be obtained. Therefore, in the cross-mixing test, it is a difficult measurement that cannot completely cope with the preparation conditions of a series of mixed plasma samples by only extending the method disclosed in Example 1.

[0232] When using the index of each feature space (28 dimensions) obtained by normalization based on the NRC method in this embodiment, a maximum discrimination accuracy of 76.5% was obtained in feature space #3. It is believed that the influence of measurement variation is also strongly involved.

[0233] Fig.25AIt is a diagram showing the clustering area of ​​the indicators (F1, F2) of Example 5. As can be seen in the figure, the FVIII deficiency cluster and the FIX deficiency cluster are close to each other and almost overlap. It can be seen that although they are independent clusters as coagulation factor deficiency groups, it is difficult to distinguish with high accuracy from the perspective of which one of the FVIII factor and the FIX factor is lacking. Although it is difficult to determine which coagulation factor is lacking, it is a feature of the existing method, i.e., the cross-mixing test, in that it can be judged as a factor deficiency type. In addition, the central part of the LA positive cluster (LA) and the coagulation factor inhibitor positive cluster (FVIII deficiency inh) overlap, and the experimental results recorded in paragraph 0121 of Patent Document 1 can be obtained, "The changes in the graph patterns of the LA positive specimen and the coagulation factor inhibitor positive specimen are similar. Therefore, the identification of the two based on the changes in the graph pattern has to be a qualitative evaluation, and it can be seen that it is difficult to judge without being an expert."

[0234] Fig.25B This is a graph showing the experimental results of adding data points to the cluster area of ​​the indicators (F1, F2) in Example 5. As can be seen in the figure, the commercially available specimens prepared for the coagulation factor inhibitor positive cluster (FVIII deficiency inh) have two types of inhibitor titers, "Mild" and "Strong", and it can be seen that the LA positive cluster (LA) is distributed in the lower left and upper right. Therefore, if a specimen of a patient with an intermediate degree between "Mild" and "Strong" is assumed, it becomes FVIII deficiency inh≈LA in this method, and it is considered that identification cannot be performed.

[0235] Fig.26 This is a diagram illustrating the index (G1, G2) of Example 5. As described above, it is difficult to separate the FVIII deficiency cluster and the FIX deficiency cluster by using only the coagulation time. Therefore, the midpoint of the line segment connecting the center of the FVIII deficiency cluster and the center of the FIX deficiency cluster in each feature space is used as the origin, and the length of the vertical line drawn from the position vector to the data P of 100% of the test specimen is used as the index. The indexes (G1, G2) corresponding to the immediate measurement and the delayed measurement are defined by the secondary formula.

[0236] [Formula 11]

[0237]

[0238] In the above formula, the center of the FVIII-deficient cluster is +0.5, and the center of the FIX-deficient cluster is -0.5. A larger value means an increased possibility that the sample is FVIII-deficient.

[0239] Fig.27AThis is a graph (display example) showing the experimental results of Example 5, which plots the immediate (F1, G1) and delayed (F2, G2) combinations of the index F calculated based on a series of coagulation waveforms obtained from one specimen and the index G calculated using the same specimen based on the data of feature space #1 and according to formula 11. Here, one specimen is represented as two points (F1, G1) and (F2, G2) and a straight line connecting them. The appearance is close to the schematic diagram of a dipole. Hereinafter, it will be referred to as a dipole. As can be seen in the figure, the clusters of the four specimen groups (FVIII deficiency, FIX deficiency, LA, FVIII deficiency inh) are formed in roughly independent areas. The upper limit of the Mahalanobis distance is set to 6σ. As mentioned above, the region of FVIII deficiency inh is relatively wide because there are two types of inhibitor titers, "Mild" and "Strong".

[0240] Fig.27B This is a graph (display example) showing the experimental results of Example 5, which plots the indicators (F1, G1) and (F2, G2) obtained from 34 specimens. Here, the indicator G is calculated based on the data for the feature space #1 and according to formula 11. It can be seen that the dipoles characterizing each specimen are separated and plotted according to each cause of the extension of the clotting time. The identification accuracy is 97%. It can be seen that "Mild" and "Strong" of the coagulation factor inhibitor positive group (FVIII deficiency inh), the LA positive group (LA), the FVIII deficiency group, and the FIX deficiency group are spatially separated and plotted. It was confirmed that, by this method, the identification accuracy is greatly improved compared with the existing identification using only the clotting time.

[0241] Next, as a specific example of a method suitable for applying the technology of the present invention to a cross-mix test, a method of distinguishing the results of a cross-mix test using feature value spaces #0 to #5 is described.

[0242] In the feature space, seven-point sequences obtained in the immediate measurement and delayed measurement of one sample are connected in ascending or descending order of the mixing rate of the sample to be tested, and the result of the connection is referred to as trajectory data. Fig.28A It is an experimental result (display example) showing the average shape of the trajectory data of each cause of prolongation of the clotting time obtained by the cross-mixing test of Example 5. As can be seen in the figure, it can be seen that the shape of the trajectory data is characteristic according to the cause of prolongation of each clotting time. For example, although the FVIII deficiency group and the coagulation factor inhibitor positive group (FVIII deficiency inh) are close in morphology (shape similarity), the coagulation factor inhibitor positive group (FVIII deficiency inh) has a large deviation from the immediate (Wait = 0h) and delayed (Wait = 2h). This situation is compared with the nature of the cross-mixing test and can be judged as a correct result.

[0243] As one of the methods to identify the cause of prolonged coagulation time from trajectory data, handwriting recognition technology such as CNN (Convolutional Neural Network) is effective. It is a method that connects the points that constitute the trajectory data using straight lines, spline interpolation, etc., generates image data, and identifies it using the same method as handwriting recognition. This method has the ability to identify many causes of prolonged coagulation time not only corresponding to the clustering of the four specimen groups treated this time, but also corresponding to the status of the patient being treated (including medication, etc.).

[0244] As another method to identify the cause of prolonged coagulation time from trajectory data, a multidimensional mixed Gaussian distribution can be used. Fig.24 As shown, it can be seen that high accuracy cannot be obtained. The main reason is that the clustering of each specimen group is defined as the interior of a 28-dimensional solid according to the aforementioned conditions, but the mathematical restriction that the solid must be Gaussian distributed no matter which plane it is sliced ​​on is insufficient for the identification of trajectory data adapted to cross-mixing tests. As is well known, in a multidimensional mixed Gaussian distribution, it must be non-degenerate, but in a two-dimensional mixed Gaussian distribution, the restriction is relaxed. Therefore, the 28-dimensional trajectory data is separated into a two-dimensional x14 space, and the product of the probability of belonging to the clustering of the specimen group obtained in each space is used to implement the identification of the cause of the extension of the coagulation time. As a result, the identification of the cause of the extension of the coagulation time can be implemented without being bound by the limitation of the continuity of the trajectory data relative to the change in the mixing rate of the specimen.

[0245] exist Fig.28B The experimental results (display example) of separating the trajectory data of each specimen in the cross-mixing test into a two-dimensional x14 space for identification are shown. Here, identification is performed based on the product of the belonging probabilities obtained in the separate spaces and the belonging probabilities of clustering to the specimen groups are normalized, and the trajectory data is plotted in four areas. In the figure, the results of feature space #4 are shown as an example. This is the result of simply expanding the method disclosed in Example 1. Fig.24 (Identification accuracy = 32.4% to 88.2%), a significant improvement was confirmed, and the identification accuracy was 100%.

[0246] The complexity of the preparation and cultivation of the specimens accompanying the cross-mixing test becomes a burden on the examiner. In addition, the additional blood sampling for the test increases the burden on the patient. Therefore, it is preferred that the specimen conditions used in the cross-mixing test (the total number of specimens prepared according to the mixing ratio and the difference in cultivation) are as small as possible.

[0247] Fig.29This is a diagram summarizing the conditions for the total number of specimens required for the investigation of the identification method in Example 5. As mentioned above, 14 (7 series of immediate and delayed) coagulation waveforms were obtained for one specimen this time. Each row represents the conditions of the coagulation waveform, and the shaded conditions are not used in the coagulation waveform analysis. In the figure, "Condition A" is the case of using all 14 coagulation waveforms, and "Condition B" represents the condition for identification using 7 coagulation waveforms excluding the shaded waveforms.

[0248] Fig.30 This is a diagram showing the relationship between the number of coagulation waveforms used in the identification of Example 5 and the worst probability (when the probability of the i-th specimen being clustered into the correct specimen group is set to Pti, it is the minimum value of Pti among all specimens). It can be considered that if the worst probability on the vertical axis is 50% or more, no identification error occurs, and if it is less than 50%, an identification error occurs. Here, the performance of the following three methods is compared: (1) identification using the indicators F1 and F2 (F Factor Method); (2) identification using the combination of indicators F1, G1, F2, and G2 (FG Factor Method); (3) identification by dividing trajectory data into two-dimensional groups (Orbit Discrimination Method). For (1) the F Factor Method, the worst probability under all conditions is less than 50%, and an identification error is bound to occur. In (2) the FG Factor Method, the identification performance is improved, and the worst probability becomes more than 50% when more than 5 coagulation waveforms are used. (3) The trajectory identification method has the best performance. If two or more coagulation waveforms are used, the worst probability becomes more than 77%.

[0249] The F-factor method uses the result of indexing the relationship between the mixing conditions of the test sample and the normal sample and the coagulation time according to the above definition. Therefore, it can be considered that its discrimination performance is in accordance with the discrimination results of the existing cross-mixing test.

[0250] The FG factor method seeks to improve the identification performance as shown in the figure by adding the shape characteristics of the coagulation waveform quantified by the NRC method of the present invention as indicators G1 and G2 in addition to the indicators F1 and F2. In addition, the amount of data to be processed is as small as 4 per specimen, and the information processing operation can be performed in a short time compared with the trajectory recognition method. Therefore, the FG factor method is used simultaneously with the existing method of identifying the cause of the extension of the coagulation time based on the relationship between the mixing ratio of the specimen and the coagulation time, and is suitable for use as auxiliary information when judging the identification result.

[0251] The trajectory identification method can obtain high identification performance by implementing the identification of the results of the cross-mixing test only by using the index of the feature space normalized and digitized by the NRC method of the present invention. At the same time, since there is no calculation of the index F based on the formula 10, it is possible to perform the same identification as that via Δx i The independent information calculation of the association with adjacent data points can flexibly respond to the selection of any mixing ratio. With these advantages, the trajectory identification method is suitable as a means to realize the fully automatic identification of cross-mixing experiments based on machine learning technology.

[0252] Fig.31 This is an example of display of the identification results of the cross-mixing test in Example 5. The relationship between the analysis unit 130, the control computer 120 and the communication interface 124, and the display of the results of the test sample on the display unit 118c Fig. 9 result.

[0253] According to Example 5, quantitative discrimination based on the probability of the LA-positive sample group, the coagulation factor inhibitor-positive sample group, the FVIII-deficient sample group, and the FIX-deficient sample group can be performed based on a series of coagulation waveforms obtained in the cross-mixing test.

[0254] [Modifications]

[0255] In addition, the present disclosure is not limited to the above-mentioned embodiments, and includes various modified examples. The above-mentioned embodiments are described in detail to explain the present disclosure in an easy-to-understand manner, but are not necessarily limited to all structures having the description. A part of the structure of a certain embodiment can be replaced with the structure of other embodiments, and the structure of other embodiments can be added to the structure of a certain embodiment. In addition, the same structure or other structures can be added, deleted, or replaced with a part of the structure of each embodiment.

[0256] For example, in the first embodiment, the cause of the extension of the coagulation time is estimated based on the characteristic quantities of WaveNor0, 1, and 2, but the present disclosure may also estimate the cause of the extension of the coagulation time based on at least two characteristic quantities of WaveNor0, 1, and 2. For example, the cause of the extension may be estimated based on the characteristic quantities of the x component and the y component of WaveNor0, or the cause of the extension may be estimated based on the characteristic quantities of the x component of WaveNor0 and the y component of WaveNor2.

[0257] Description of Reference Numerals

[0258] 100: automatic analyzer, 101: sample dispensing mechanism, 102: sample disk, 103: sample container, 103a: sample, 104: reaction container, 105: sample syringe pump, 106: reagent dispensing mechanism, 107: reagent disk, 108: reagent container, 108a: reagent, 109: reagent heating mechanism, 110: reagent syringe pump, 111: reaction container storage unit, 112: reaction container transport mechanism, 113: detection group 114: reaction container setting part, 115: light source, 116: detection part (photo sensor), 117: reaction container disposal part, 118: operation computer, 118a: mouse, 118b: keyboard, 118c: display part, 119: storage part, 120: control computer, 121: A / D converter, 122: incubator, 123: printer, 124: communication interface, 125: analysis computer, 130: analysis part.

Claims

1. A method for estimating the cause of prolonged coagulation time, characterized in that: have: Acquiring a coagulation waveform indicating a change in the amount of light over time caused by a coagulation reaction of a reaction solution, wherein the reaction solution is generated by mixing a test sample consisting of plasma obtained by separating blood obtained from a subject and a reagent; Obtaining a first waveform by performing front-to-back differential processing on the coagulation waveform; Obtaining a first fitting waveform by fitting the coagulation waveform; Obtaining a second fitting waveform by fitting the first waveform; Obtaining a third waveform by performing front-to-back difference processing on the second fitting waveform; acquiring a first normalized waveform, a second normalized waveform, and a third normalized waveform by normalizing the light quantity axis and the time axis of each of the first fitting waveform, the second fitting waveform, and the third waveform; extracting a feature quantity from each of the first normalized waveform, the second normalized waveform, and the third normalized waveform; and The cause of prolonged coagulation time of the sample is estimated based on known feature quantities extracted from a sample group with known causes of prolonged coagulation time and the feature quantities extracted from each of the first normalized waveform, the second normalized waveform, and the third normalized waveform.

2. The method for estimating the cause of prolonged coagulation time according to claim 1, wherein: The normalization is based on the numerical value obtained by fitting the coagulation waveform.

3. The method for estimating the cause of prolonged coagulation time according to claim 1, wherein: Estimating the cause of the prolonged coagulation time includes estimating that the cause of the prolonged coagulation time of the test subject is coagulation factor VIII deficiency, coagulation factor IX deficiency, or lupus anticoagulant LA positivity.

4. The method for estimating the cause of prolonged coagulation time according to claim 3, wherein: The blood coagulation factor VIII deficiency is a state in which the activity value of the blood coagulation factor VIII is less than 1%, and the blood coagulation factor IX deficiency is a state in which the activity value of the blood coagulation factor IX is less than 1%.

5. The method for estimating the cause of prolonged coagulation time according to claim 1, wherein: The method for estimating the cause of the prolonged coagulation time further comprises: before the fitting process on the coagulation waveform, performing a first filtering process for removing noise from the coagulation waveform; as well as Before the fitting process on the first waveform, a second filtering process for removing noise from the first waveform is performed.

6. The method for estimating the cause of prolonged coagulation time according to claim 5, wherein: In the second filtering process, a moving average process is performed using a response time constant which is one of the parameters obtained by the fitting process on the coagulation waveform.

7. The method for estimating the cause of prolonged coagulation time according to claim 1, wherein: The method for estimating the cause of the prolonged coagulation time further comprises: The estimated reason for the extension is displayed on the display unit.

8. The method for estimating the cause of prolonged coagulation time according to claim 1, wherein: The method for estimating the cause of the prolonged coagulation time further comprises: creating a map showing a relationship between known feature quantities extracted from the sample group for which the cause of prolonged coagulation time is known and the cause of prolonged coagulation time; and The map and information in which the feature amount of the subject is plotted on the map are displayed on a display unit.

9. The method for estimating the cause of prolonged coagulation time according to claim 1, wherein: The method for estimating the cause of the prolonged coagulation time further comprises: The identification and estimation result based on the estimated cause of the prolonged coagulation time of the test sample is displayed on the display unit.

10. An information processing device capable of communicating with an automatic analysis device, the automatic analysis device acquiring a coagulation waveform representing a change in light intensity over time caused by a coagulation reaction of a reaction solution, wherein: The reaction solution is generated by mixing a test sample consisting of plasma separated from blood obtained from a test subject and a reagent. The information processing device is characterized in that The information processing device includes a computer system having a processor and a memory. The computer system performs the following processing: obtaining the coagulation waveform from the automatic analysis device; Obtaining a first waveform by performing front-to-back differential processing on the coagulation waveform; Obtaining a first fitting waveform by fitting the coagulation waveform, and obtaining a second fitting waveform by fitting the first waveform; Obtaining a third waveform by performing front-to-back difference processing on the second fitting waveform; acquiring a first normalized waveform, a second normalized waveform, and a third normalized waveform by normalizing the light quantity axis and the time axis of each of the first fitting waveform, the second fitting waveform, and the third waveform; extracting a feature quantity from each of the first normalized waveform, the second normalized waveform, and the third normalized waveform; as well as The cause of prolonged coagulation time of the sample is estimated based on known feature quantities extracted from a sample group with known causes of prolonged coagulation time and the feature quantities extracted from each of the first normalized waveform, the second normalized waveform, and the third normalized waveform.

11. The information processing device according to claim 10, wherein: The computer system performs normalization based on a numerical value obtained by a fitting process on the coagulation waveform.

12. The information processing device according to claim 10, wherein: The process of estimating the cause of the prolonged coagulation time includes a process of estimating that the cause of the prolonged coagulation time of the test sample is a deficiency of coagulation factor VIII, a deficiency of coagulation factor IX, or a positive lupus anticoagulant LA.

13. The information processing device according to claim 12, wherein: The blood coagulation factor VIII deficiency is a state in which the activity value of the blood coagulation factor VIII is less than 1%, and the blood coagulation factor IX deficiency is a state in which the activity value of the blood coagulation factor IX is less than 1%.

14. The information processing device according to claim 10, wherein: The computer system further performs the following processing: before the fitting process on the coagulation waveform, performing a first filtering process of removing noise from the coagulation waveform; and Before the fitting process on the first waveform, a second filtering process is performed to remove noise from the first waveform.

15. The information processing device according to claim 14, wherein: The computer system executes a moving average process in the second filtering process using a response time constant that is one of the parameters obtained by the fitting process on the coagulation waveform.

16. The information processing device according to claim 10, wherein: The information processing device further includes a display unit that displays the estimated reason for the extension.

17. The information processing device according to claim 10, wherein: The computer system further performs the following processing: creating a map showing a relationship between known feature quantities extracted from the sample group for which the cause of prolonged coagulation time is known and the cause of prolonged coagulation time; and The map and information in which the feature amount of the subject is plotted on the map are displayed on a display unit.

18. The information processing device according to claim 10, wherein: The information processing apparatus further includes a display unit that displays a result of identification and estimation based on the estimated cause of prolongation of the blood coagulation time of the sample.

19. A method for estimating a cause of prolonged coagulation time of a test specimen, the method comprising estimating a cause of prolonged coagulation time of the test specimen based on a coagulation waveform indicating a change in light intensity over time caused by a coagulation reaction of a reaction solution, wherein: The reaction solution is generated by mixing the test sample consisting of plasma separated from blood taken from a test subject and a reagent. The method for estimating the cause of prolonged coagulation time is characterized by comprising: making the data length of the coagulation waveform data consistent; Using a neural network to estimate the causes of prolongation of multiple coagulation times using the coagulation waveform data with the same data length; as well as The cause of the prolonged coagulation time of the test sample is presented based on the estimated probability of belonging to the cause of the prolonged coagulation time.

20. The method for estimating the cause of prolonged coagulation time according to claim 19, wherein: The neural network in the estimation method is a neural network having 16 or more nodes and 3 or more hidden layers.

21. A method for estimating the cause of prolonged coagulation time, characterized in that: have: obtaining a coagulation waveform indicating a change in the amount of light over time caused by a coagulation reaction of a reaction solution, wherein the reaction solution is generated by mixing a test sample consisting of plasma obtained by separating blood obtained from a subject and a reagent, Obtaining a first waveform by performing front-to-back differential processing on the coagulation waveform; Obtaining a first fitting waveform by fitting the coagulation waveform; Obtaining a second fitting waveform by fitting the first waveform; Obtaining a third waveform by performing front-to-back difference processing on the second fitting waveform; acquiring a first normalized waveform, a second normalized waveform, and a third normalized waveform by normalizing the light quantity axis and the time axis of each of the first fitting waveform, the second fitting waveform, and the third waveform; extracting a feature quantity from at least one of the first normalized waveform, the second normalized waveform, the third normalized waveform, and waveform data obtained by performing a nonlinear operation on these waveforms; and The cause of prolongation of the coagulation time of the subject is estimated based on the distribution information of the feature amount for each prolongation cause of the coagulation time prepared in advance and the feature amount extracted from the coagulation waveform of the subject.

22. A method for estimating the cause of prolonged coagulation time, characterized in that: have: preparing a series of sample specimens in which plasma separated from blood obtained from a subject and normal plasma are mixed at a plurality of mixing ratios; acquiring a series of coagulation waveforms representing changes in the amount of light over time caused by a coagulation reaction of a reaction solution generated by mixing the series of sample specimens and reagents; Obtaining a series of first waveforms by respectively performing front-to-back differential processing on the series of coagulation waveforms; Obtaining a series of first fitting waveforms by fitting the series of coagulation waveforms respectively; Obtaining a series of second fitting waveforms by fitting the series of first waveforms respectively; Obtaining a series of third waveforms by respectively performing forward and backward difference processing on the series of second fitting waveforms; Normalizing the light quantity axis and the time axis of each of the series of first fitting waveforms, the series of second fitting waveforms, and the series of third waveforms to obtain a series of first normalized waveforms, a series of second normalized waveforms, and a series of third normalized waveforms; extracting a feature quantity from at least one of the series of first normalized waveforms, the series of second normalized waveforms, the series of third normalized waveforms, and a series of waveform data obtained by performing a nonlinear operation on these; as well as The cause of the prolonged coagulation time of the sample specimen is estimated using at least one of the series of feature quantities extracted from the series of coagulation waveforms.

23. The method for estimating the cause of prolonged coagulation time according to claim 22, wherein: The feature quantities extracted from the series of waveform data are a combination of index G1 and index F1 (F1, G1) and a combination of index G2 and index F2 (F2, G2), The index G1 is calculated based on the feature quantities including a plurality of components obtained from the instantaneous measurement of the coagulation waveform of a sample specimen having a mixing ratio of 100% of plasma separated from the blood obtained from the subject among the series of sample specimens, and the index F1 is calculated based on the series of coagulation times instantaneously measured for the series of sample specimens. The index G2 is calculated based on the position of the feature quantity space by plotting the feature quantity including the plurality of components obtained from the delayed-measured coagulation waveform of the sample specimen having a mixing rate of 100% of the plasma, and the index F2 is calculated based on a series of coagulation times of the delayed-measured series of the sample specimens, The cause of the prolonged coagulation time of the sample specimen is estimated using the combination (F1, G1) and the combination (F2, G2).

24. The method for estimating the cause of prolonged coagulation time according to claim 22, The feature quantities extracted from the series of waveform data are first trajectory data and second trajectory data. in, A series of feature quantities including a plurality of components obtained from a series of coagulation waveforms measured in real time from the series of sample specimens are plotted in a feature quantity space representing the plurality of components, and the series of feature quantities in the feature quantity space are connected in ascending or descending order of the mixing rate of the plasma, thereby obtaining the first trajectory data. wherein a series of feature quantities including the plurality of components obtained from a series of coagulation waveforms of delayed measurements of the series of sample specimens are plotted in the feature quantity space, and the series of feature quantities in the feature quantity space are connected in ascending or descending order of the mixing rate of the plasma, thereby obtaining the second trajectory data; The cause of the prolonged coagulation time of the sample specimen is estimated using the first trajectory data and the second trajectory data.

25. An information processing device capable of communicating with an automatic analysis device, the automatic analysis device acquiring a coagulation waveform representing a change in light intensity over time caused by a coagulation reaction of a reaction solution, wherein: The reaction solution is generated by mixing a test sample consisting of plasma separated from blood obtained from a test subject and a reagent. The information processing device is characterized in that The information processing device includes a computer system having a processor and a memory. The computer system performs the following processing: obtaining the coagulation waveform from the automatic analysis device; Obtaining a first waveform by performing front-to-back differential processing on the coagulation waveform; Obtaining a first fitting waveform by fitting the coagulation waveform; Obtaining a second fitting waveform by fitting the first waveform; Obtaining a third waveform by performing front-to-back difference processing on the second fitting waveform; acquiring a first normalized waveform, a second normalized waveform, and a third normalized waveform by normalizing the light quantity axis and the time axis of each of the first fitting waveform, the second fitting waveform, and the third waveform; extracting a feature quantity from at least one of the first normalized waveform, the second normalized waveform, the third normalized waveform, and waveform data obtained by performing a nonlinear operation on these waveforms; as well as The cause of prolongation of the coagulation time of the subject is estimated based on the distribution information of the feature amount for each prolongation cause of the coagulation time prepared in advance and the feature amount extracted from the coagulation waveform of the subject.

26. The information processing device according to claim 25, wherein: The information processing device further performs the following processing: creating a graph showing a relationship between the distribution information of the feature quantity for each of the causes of prolongation of the coagulation time prepared in advance and the feature quantity extracted from the coagulation waveform of the subject; as well as The graph and the estimated result of the cause of the prolonged coagulation time of the test sample are displayed on a display unit.

27. An information processing device capable of communicating with an automatic analyzing device, the automatic analyzing device preparing a series of sample specimens of plasma obtained by separating blood from a subject and normal plasma mixed at a plurality of mixing ratios, obtaining a series of coagulation waveforms representing changes in light intensity over time caused by a coagulation reaction of a reaction solution generated by mixing the series of sample specimens and a reagent, The information processing device is characterized in that The information processing device includes a computer system having a processor and a memory. The computer system performs the following processing: Obtaining the series of coagulation waveforms from the automatic analysis device; Obtaining a series of first waveforms by respectively performing front-to-back differential processing on the series of coagulation waveforms; Obtaining a series of first fitting waveforms by fitting the series of coagulation waveforms respectively; Obtaining a series of second fitting waveforms by fitting the series of first waveforms respectively; Obtaining a series of third waveforms by respectively performing forward and backward difference processing on the series of second fitting waveforms; Normalizing the light quantity axis and the time axis of each of the series of first fitting waveforms, the series of second fitting waveforms, and the series of third waveforms to obtain a series of first normalized waveforms, a series of second normalized waveforms, and a series of third normalized waveforms; extracting a feature quantity from at least one of the series of first normalized waveforms, the series of second normalized waveforms, the series of third normalized waveforms, and a series of waveform data obtained by performing a nonlinear operation on these; as well as The cause of the prolonged coagulation time of the sample specimen is estimated using at least one of the series of feature quantities extracted from the series of coagulation waveforms.

28. The information processing device according to claim 27, wherein: The feature quantities extracted from the series of waveform data are a combination of index G1 and index F1 (F1, G1) and a combination of index G2 and index F2 (F2, G2), The index G1 is calculated based on the feature quantities including a plurality of components obtained from the instantaneous measurement of the coagulation waveform of a sample specimen having a mixing ratio of 100% of plasma separated from the blood obtained from the subject among the series of sample specimens, and the index F1 is calculated based on the series of coagulation times instantaneously measured for the series of sample specimens. The index G2 is calculated based on the position of the feature quantity space by plotting the feature quantity including the plurality of components obtained from the delayed-measured coagulation waveform of the sample specimen having a mixing rate of 100% of the plasma, and the index F2 is calculated based on a series of coagulation times of the delayed-measured series of the sample specimens, The process of estimating the cause of the prolongation is a process of estimating the cause of the prolongation of the blood coagulation time of the sample specimen using the combination (F1, G1) and the combination (F2, G2). The information processing device further performs the following processing: creating a graph showing the relationship between the index G1 and the index F1, and the relationship between the index G2 and the index F2; and The graph and the estimated result of the cause of the prolonged coagulation time of the sample specimen are displayed on a display unit.

29. The information processing device according to claim 27, wherein: The feature quantities extracted from the series of waveform data are first trajectory data and second trajectory data. wherein a series of feature quantities including a plurality of components obtained from a series of coagulation waveforms measured in real time from the series of sample specimens are plotted in a feature quantity space representing the plurality of components, and the series of feature quantities in the feature quantity space are connected in ascending or descending order of the mixing rate of the plasma, thereby obtaining the first trajectory data; A series of feature quantities including the plurality of components obtained from a series of coagulation waveforms of the series of sample specimens by delayed measurement are plotted in the feature quantity space, and the series of feature quantities in the feature quantity space are connected in ascending or descending order of the mixing rate of the plasma, thereby obtaining second trajectory data. The process of estimating the cause of the prolongation is a process of estimating the cause of the prolongation of the coagulation time of the sample specimen using the first trajectory data and the second trajectory data. The information processing device further performs the following processing: creating a graph representing the first trajectory data and the second trajectory data; and The graph and the estimated result of the cause of the prolonged coagulation time of the sample specimen are displayed on a display unit.

Images