This application claims priority to Japanese Patent Application No. 2020-171206, filed on Oct. 9, 2020, the entire content of which is incorporated herein by reference.
The present specification discloses a blood specimen analysis method, an analyzer, and an analysis program.
In a blood coagulation test, when the blood coagulation time has been prolonged, identification of the cause is performed by a medical worker. As a method for supporting identification of a cause of blood coagulation time prolongation, U.S. Pat. No. 6,321,164 discloses a method for predicting, by use of a neural network, coagulation factor deficiency, contamination of heparin into the specimen, and the presence of lupus anticoagulant. In U.S. Pat. No. 6,321,164, a PT or APTT time-dependent optical profile is obtained; and one or more predictor variables selected from a minimum value of a first derivative of the optical profile, a time index of the minimum value of the first derivative, a minimum value of a second derivative of the optical profile, a time index of the minimum value of the second derivative, a maximum value of the second derivative, a time index of the maximum value of the second derivative, the overall change of transmittance during reaction, a coagulation time, the slope of the optical profile before clot formation, and the slope of the optical profile after the clot formation, are used as input values to the neural network.
In the method described in U.S. Pat. No. 6,321,164, the input values to the neural network are limited to specific predictor variables. Therefore, when changes in the predictor variables with respect to a plurality of prolongation causes are similar, prediction is difficult, and thus, improvement of prediction accuracy has been desired.
The present invention addresses providing a blood specimen analysis method, an analyzer, and an analysis program that can predict prolongation causes with higher accuracy than before.
An embodiment disclosed in the present specification relates to an analysis method for a blood specimen. The analysis method includes: obtaining a data group including a plurality of data forming a blood coagulation curve or a differential curve thereof; inputting the data group into a deep learning algorithm; and outputting, on the basis of a result obtained from the deep learning algorithm, information regarding a cause of prolongation of blood coagulation time of the blood specimen.
Another embodiment disclosed in the present specification relates to an analyzer (1) for a blood specimen. The analyzer (1) includes a measurement unit (2) and a controller (201). The measurement unit (2) is configured to prepare a measurement sample that contains the blood specimen and a coagulation time measurement reagent, and configured to output a plurality of pieces of detection information forming a blood coagulation curve, on the basis of the measurement sample. The controller (201) is configured to: obtain a data group including a plurality of data forming the blood coagulation curve or a differential curve thereof, on the basis of the plurality of pieces of detection information; input the data group into a deep learning algorithm; and output, on the basis of a result obtained from the deep learning algorithm, information regarding a cause of prolongation of blood coagulation time of the blood specimen.
Another embodiment disclosed in the present specification relates to an analysis program (202b) for a blood specimen. The analysis program (202b) is configured to cause, when executed by a computer, the computer to execute the steps of: obtaining a data group including a plurality of data forming a blood coagulation curve or a differential curve thereof; inputting the data group into a deep learning algorithm; and outputting, on the basis of a result obtained from the deep learning algorithm, information regarding a cause of prolongation of blood coagulation time of the blood specimen.
According to the present invention, it is possible to predict a prolongation cause with higher accuracy than before.
With reference to
The analyzer 1 is an apparatus in which light is applied to a measurement sample prepared by adding a coagulation measurement reagent to a blood specimen, transmitted light of the light applied to the measurement sample is detected, and the blood specimen is analyzed on the basis of the detected light.
The measurement unit 2 of the analyzer 1 includes a controller 201, a storage 202, a light applicator 10, a sample preparation part 20, a detector 230, an input interface (I/F) 206, an output interface (I/F) 207, a communication interface (I/F) 208, and a bus 209.
The controller 201 includes an arithmetic processing device such as a CPU (Central Processing Unit) or an FPGA (field-programmable gate array).
The storage 202 stores: a measurement program 202a for controlling measurement operation performed by the measurement unit 2; an analysis program 202b; an algorithm database (DB) 202c storing one or a plurality of deep learning algorithms 60; a reference value/threshold database (DB) 202d storing a reference value for a blood coagulation time of each blood coagulation parameter, and a threshold for a probability that a cause candidate for blood coagulation time prolongation is the cause of blood coagulation time prolongation; and an additional test database (DB) 202e storing information of additional tests.
The input interface 206 receives input information inputted by an operator through an input unit 411 of the display 4, and transmits the input information to the controller 201 or the storage 202.
The output interface 207 transmits, to an output unit 412 of the display 4, output information outputted by the controller 201.
The communication interface 208 communicably connects the measurement unit 2 to a network 99. The connection may be wired connection or wireless connection.
Signal transmission in the measurement unit 2 is performed via the bus 209.
Each of the five light sources 320 is implemented as an LED. In general, the life of an LED is several tens of times as long as that of a halogen lamp. Therefore, when compared with a configuration in which a wide-band light source such as a halogen lamp and a rotary filter are used, a light applicator 10 that is smaller and that has a longer life can be configured. In addition, since LEDs can be provided individually for respective wavelengths, emission spectra and emission intensities of the respective light sources 320 can be individually optimized.
The light sources 320 include a first light source 321, a second light source 322, and a third light source 323.
In the configuration example in
In the configuration example in
In the configuration example in
The optical fiber parts 330 are provided so as to correspond to the respective light sources 320. The five optical fiber parts 330 are implemented as optical fiber parts 330a, 330b, 330c, 330d, and 330e that are individually provided for the respective light sources 320 such that lights from the first light source 321, the second light source 322, the third light source 323, the fourth light source 324, and the fifth light source 325 enter from the respective light entry ends 331.
In the configuration example in
In addition, at the common light outputting end 332, light can be outputted in a state where the lights having the respective wavelengths are uniformly distributed. Therefore, even when lights having the respective wavelengths are outputted from the common light outputting end 332, biased distribution of lights of the respective wavelengths can be suppressed.
In the configuration example in
In the configuration example in
The uniformization member 350 is disposed at each of the two outlets 311 provided in the housing 310. As for each uniformization member 350, a light entry face 351 is opposed to the corresponding light outputting end 332 of the optical fiber part 330, and a light outputting face 352 is disposed on the exit side at the outlet 311. Accordingly, light of which the intensity distribution has been made uniform through the uniformization member 350 is outputted from each outlet 311. The uniformization member 350 is configured such that, for example, light having entered from the light entry face 351 is reflected multiple times to be outputted from the light outputting face 352.
In a case where the intensity distribution of lights having the respective wavelengths is sufficiently made uniform at the light outputting end 332 of the optical fiber part 330, the uniformization member 350 need not necessarily be provided.
The holding member 340 of the light applicator 10 holds the five light sources 320. Therefore, the five light sources 320 are supported by the common holding member 340. The holding member 340 is made of metal such as aluminum, for example, and is formed in a prism shape. In the configuration example in
The five light source holders 341 are disposed so as to be linearly arranged along a direction orthogonal to the outputting direction of the light of each light source 320. As for the light sources 320, the fourth light source 324 is disposed at the center, the fifth light source 325 and the second light source 322 are disposed on both sides of the fourth light source 324, and the first light source 321 and the third light source 323 are disposed on the outermost sides.
In the configuration example in
In the configuration example in
The light applicator 10 may be provided with a member for condensing the light from each light source 320 on the light entry end 331 of the optical fiber part 330, and a member for adjusting spectrum characteristics such as the center wavelength or the half-width of light entering the light entry end 331.
For example, the light applicator 10 further includes optical band-pass filters 360 that each allow only light in a predetermined wavelength band to be transmitted therethrough. Each optical band-pass filter 360 has a disk-like shape, and allows, out of the light applied to one surface thereof, only light in a predetermined wavelength band to be transmitted to the other surface. The holding member 340 holds each optical band-pass filter 360 at a position between a light source 320 and a light entry end 331 of a corresponding optical fiber part 330. Accordingly, the center wavelength or half-width of the light outputted from the light source 320 can be adjusted so as to have characteristics appropriate for measurement, to be caused to enter the light entry end 331. As a result, measurement accuracy is improved. Although there are cases where individual differences are present in the light sources 320 and the center wavelength and half-width are different, it is possible to absorb influence of the individual differences of each light source 320 by the optical band-pass filter 360, thereby ensuring a stable measurement result.
Here, an example in which LEDs are used as the light sources has been described. However, for example, a halogen lamp may be used as a light source, the light thereof may be split, by band-pass filters or the like, into lights having the first wavelength to the fifth wavelength, and the respective lights may be applied to the measurement sample.
With reference back to
As the light applicator 10, the sample preparation part 20, and the detector 230, the light applicator, the sample preparation part, and the detector described in U.S. Pat. No. 10,048,249 can be used, for example. The content of U.S. Pat. No. 10,048,249 is incorporated by reference in the present specification.
In step S11, the controller 201 obtains, for each blood specimen, information of a blood coagulation parameter (analysis item) ordered for the blood specimen. As the information of the blood coagulation parameter, information inputted by the operator through the display 4 may be received. Alternatively, the information of the blood coagulation parameter may be obtained via a network from, for example, an electronic health record system of a medical facility. The blood specimen and the information of the blood coagulation parameter can be associated with each other by means of an identifier issued at the time of request of a test, for example.
The blood coagulation parameter (analysis item) targeted by the analyzer 1 includes at least one type selected from the group consisting of: activated partial thromboplastin time (hereinafter, this may be abbreviated as “APTT”), prothrombin time (hereinafter, this may be abbreviated as “PT”), thrombo test, fibrinogen, factor II activity, factor V activity, factor VII activity, factor VIII activity, factor IX activity, factor X activity, factor XI activity, factor XII activity, and whole blood coagulation time. Preferably, the blood coagulation parameter can include at least one type selected from the group consisting of APTT and PT.
In step S12, the controller 201 controls the measurement unit 2 so as to dispense a blood specimen into a container 15. At this time, a buffer or the like for diluting the blood specimen may be dispensed into the container 15.
In step S13, the controller 201 controls the sample preparation part 20 of the measurement unit 2 so as to dispense a coagulation time measurement reagent corresponding to each blood specimen to prepare a measurement sample. At this time, when a coagulation activation reagent needs to be added, the coagulation activation reagent is also dispensed. The coagulation time measurement reagent can be selected as appropriate in accordance with each blood coagulation parameter. As the coagulation time measurement reagent, a commercially available reagent can be used.
For example, when APTT is to be measured, an APTT measurement test reagent that can contain: an activator such as silica, ellagic acid, or celite; an animal-derived, plant-derived, or artificially-synthesized phospholipid; and the like, can be used as the coagulation time measurement reagent. Examples thereof can include Thrombocheck APTT series manufactured by Sysmex Corporation, Coagpia (registered trademark) APTT-N, etc., of Sekisui Medical Co., Ltd., and Data-Fi-APTT, etc., of Siemens Healthcare Diagnostics Products GmbH.
The coagulation activation reagent is a reagent that can supply calcium ions. According to the International Committee for Standardization in Hematology, the coagulation activation reagent is a 20 mM calcium chloride solution.
For measurement of PT, a PT measurement test reagent that contains thrombin can be used as the coagulation time measurement reagent. Examples thereof can include Thrombocheck PT series manufactured by Sysmex Corporation, Coagpia (registered trademark) PT series by Sekisui Medical Co., Ltd., and the like. The PT measurement test reagent contains calcium ions necessary for activation of coagulation in general.
In step S14, the controller 201 controls the light applicator 10 so as to start application of light to the container 15 in which the measurement sample has been prepared in step S13. The light receiver 11 continually outputs, as detection information, electric signals (digital data) in accordance with the intensity of light received through the container 15.
In step S21, the controller 201 obtains a data group including a plurality of data forming a blood coagulation curve. Specifically, the controller 201 arrays, in time series, a plurality of data (digital data) according to the received light intensity outputted from the light receiver 11, and stores the plurality of data into the storage 202.
The controller 201 obtains detection information from the light receiver 11 over time, for example, every 0.1 seconds to 0.5 seconds, and preferably every 0.1 seconds, and stores the detection information into the storage 202. The controller 201 stores, into the storage 202, the plurality of data from the time point when, for example, the blood specimen and the predetermined coagulation time measurement reagent have been added. Generally, after the blood specimen and the coagulation time measurement reagent have been added, a coagulation activation reagent is added. In this case, the controller 201 starts storing the plurality of data from the time point when the coagulation activation reagent has been added. Alternatively, when the coagulation activation reagent has been mixed in the coagulation time measurement reagent, the controller 201 starts storing the plurality of data at the time point when the blood specimen and the coagulation time measurement reagent have been mixed. The controller 201 ends storing the plurality of data at the time point when change in the size of the plurality of data obtained over time is no longer observed. The timings of starting and ending the storing of the plurality of data is not limited thereto. For example, the storing may be started at the time point when light application to the container 15 has been started, and the storing may be ended after a predetermined time (e.g., after 120 seconds or 180 seconds) from the start of the light application.
The data group stored by the controller 201 into the storage 202 may be data obtained by removing a part of the plurality of data outputted from the light receiver 11. As to the removal of a part of data, data of a certain section, such as immediately after the start of the light application or immediately before the end of the light application, may be removed, or data may be removed at a predetermined frequency. Examples of the predetermined frequency can include, for example: when data has been obtained a predetermined number of times, data of the next time is removed; data at even-number times is removed; when data has been obtained a predetermined number of times, data corresponding to a predetermined number of times is removed; and the like.
The controller 201 may arrange the respective data in the data group to be stored in the storage 202, in time series or in an order other than time series, such as according to the intensity of detected light.
The plurality of data stored in the storage 202 forms a blood coagulation curve. The blood coagulation curve is described in more detail with reference to
In the above, an example in which the light applied to the measurement sample has one wavelength has been described. However, the light applied to the measurement sample may have a plurality of wavelengths. For example, light having a first wavelength due to the first light source 321 may be applied to the measurement sample to obtain a first plurality of data; further, light having a second wavelength due to the second light source 322 may be applied to the same measurement sample to obtain a second plurality of data; further, light having a third wavelength due to the third light source 323 may be applied to the same measurement sample to obtain a third plurality of data; further, light having a fourth wavelength due to the fourth light source 324 may be applied to the same measurement sample to obtain a fourth plurality of data; and further, light having a fifth wavelength due to the fifth light source 325 may be further applied to the same measurement sample to obtain a fifth plurality of data. In this case, the first plurality of data, the second plurality of data, the third plurality of data, the fourth plurality of data, and the fifth plurality of data each form a blood coagulation curve.
In the above embodiment, a plurality of data have been obtained on the basis of the transmitted light intensity. However, the plurality of data can be obtained by any of: an optical measurement method such as of a scattered light type or a transmitted light type; a physical method that uses magnetism and measures viscosity at the time of fibrin deposition; and a dry hematology method. In the case of an optical measurement method, the plurality of data are indicated by signals representing the light amount of transmitted light, scattered light, and the like of the light applied to the measurement sample. In the case of a physical method, the plurality of data are indicated by a signal representing the amplitude of vibration of a steel ball according to the viscosity of the measurement sample.
Next, in step S22 shown in
Normalization means expressing each data included in the data group, as a relative value so as to be shown between 0% (L1: baseline) and 100% (L2) of the vertical axis of the blood coagulation curve in
When a data group has been obtained by using light having a plurality of wavelengths, the data group expressed as relative values is generated for each wavelength.
In the normalized coagulation curve, the coagulation time can be set as a time point at which the change amount (dH) of the transmitted light intensity is, for example, 30%, 40%, 50%, or 60%. In a preferable embodiment, the coagulation time is the time at which the change amount (dH) of the transmitted light intensity is 50% (L3).
When differential processing is performed on a data group including a plurality of data, the plurality of data included in the data group form a differential curve. The differential curve includes a first-order differential coagulation curve and a second-order differential coagulation curve, for example.
When a plurality of data groups have been obtained by using light having a plurality of wavelengths, the first-order differential coagulation curve is generated for each wavelength.
When a plurality of data groups have been obtained by using light having a plurality of wavelengths, the second-order differential coagulation curve is generated for each wavelength.
In step S23 shown in
In the present embodiment, step S22 can be omitted. In this case, for example, on the basis of the blood coagulation curve in
In step S24 shown in
When the coagulation time exceeds the reference value (“YES” in step S24), the controller 201 advances to step S25 shown in
In the following, a method for generating a deep learning algorithm 60 (a method for training a deep learning algorithm 50), and the data group to be inputted to the deep learning algorithm 50 and the deep learning algorithm 60 are described.
i. Training of Deep Learning Algorithm
The deep learning algorithm is not limited as long as the algorithm has a neural network structure. A convolution neural network, a full connect neural network, and a combination of these can be included.
As training data for training the deep learning algorithm 50, a data group including a plurality of data obtained from a blood specimen for which the cause of prolongation of blood coagulation time is known is used. The data group is generated according to the method described in steps S21 and S22 shown in
Training of the deep learning algorithm 50 may be performed for each blood coagulation parameter, or a plurality of blood coagulation parameters, such as a first blood coagulation parameter and a second blood coagulation parameter, may be combined. For example, the deep learning algorithm 50 may be trained by using only a data group including a plurality of data, obtained through APTT measurement, which have been obtained from a blood specimen for which the cause of prolongation of blood coagulation time is known. In this case, the trained deep learning algorithm 60 becomes a deep learning algorithm 60 for performing analysis based on the data group including the plurality of data derived from APTT. The deep learning algorithm 50 may be trained by using only a data group including a plurality of data, obtained through PT measurement, which have been obtained from a blood specimen for which the cause of prolongation of blood coagulation time is known. In this case, the trained deep learning algorithm 60 becomes a deep learning algorithm 60 that performs analysis on the basis of the data group including the plurality of data derived from PT. Further, the deep learning algorithm 50 may be trained by using a data group including a plurality of data obtained through measurement of APTT as the first blood coagulation parameter, and a data group including a plurality of data obtained through measurement of PT as the second blood coagulation parameter. In this case, the trained deep learning algorithm 60 becomes a deep learning algorithm 60 that performs analysis on the basis of the data group including the plurality of data obtained through APTT measurement and the data group including the plurality of data derived from PT.
When a plurality of blood coagulation parameters for training are combined to produce the first training data, as shown in
As shown in
The first training data and the second training data are generated from each of a plurality of blood specimens for each of which the cause of prolongation of blood coagulation time is known, and are used for training the deep learning algorithm 50.
ii. Analysis by Trained Deep Learning Algorithm
A data group including a plurality of data, which form a blood coagulation curve and which have been obtained from a blood specimen to be analyzed, is inputted as analysis data to an input layer 60a of the trained deep learning algorithm 60 in
Preferably, the analysis data to be inputted is a data group including a plurality of data regarding the same blood coagulation parameter and having the same configuration as those of the first training data used when training the deep learning algorithm 60. For example, as shown in
When there are a plurality of deep learning algorithms 60 corresponding to blood coagulation parameters (analysis items), a deep learning algorithm 60 to which inputting is performed may be selected from the plurality of deep learning algorithms in accordance with the kind of the blood coagulation parameter.
When the cause of prolongation is to be predicted on the basis of a plurality of blood coagulation parameters, as shown in
The deep learning algorithm 60 outputs a result (in
Here, the label indicating the cause of the prolongation of the blood coagulation time and inputted as the second training data, and the label indicating the cause of the prolongation of the blood coagulation time and outputted as a result may be character information or a label value.
The cause of the prolongation of the blood coagulation time outputted from the output layer 60b of the deep learning algorithm 60 can include at least one selected from the group consisting of: liver disease; disseminated intravascular coagulation (hereinafter, this may be abbreviated as “DIC”); vitamin K (hereinafter, this may be abbreviated as “VK”) deficiency; hemorrhaging; decrease, deficiency, or dysfunction of a coagulation factor; presence of a coagulation factor inhibitor; presence of lupus anticoagulant (hereinafter, this may be abbreviated as “LA”); use of an anticoagulant drug; presence of an abnormal protein such as in macroglobulinemia; and a cause derived from a blood collection technique.
Decrease, deficiency, or dysfunction of a coagulation factor can include decrease, deficiency, or dysfunction of at least one selected from the group consisting of fibrinogen (hereinafter, this may be abbreviated as “Fbg”), factor II (hereinafter, this may be abbreviated as “FII”), factor V (hereinafter, this may be abbreviated as “FV”), factor VII (hereinafter, this may be abbreviated as “FVII”), factor VIII (hereinafter, this may be abbreviated as “FVIII”), von Willebrand factor (hereinafter, this may be abbreviated as “VWF”), factor IX (hereinafter, this may be abbreviated as “FIX”), factor X (hereinafter, this may be abbreviated as “FX”), factor XI (hereinafter, this may be abbreviated as, “FXI”), factor XII (hereinafter, this may be abbreviated as “FXII”), HMWK (High Molecular Weight Kininogen), and prekallikrein.
The coagulation factor inhibitor can include at least one selected from the group consisting of a factor V inhibitor, a factor VIII inhibitor, a von Willebrand factor inhibitor, and a factor IX inhibitor.
The anticoagulant drug is also referred to as an antithrombotic drug. The anticoagulant drug can include: a coumarin-based drug such as warfarin potassium; heparin; a synthetic Xa inhibitor such as fondaparinux sodium; an oral direct Xa inhibitor such as edoxaban tosilate hydrate and apixaban; an oral thrombin direct inhibitor such as dabigatran etexilate methanesulfonate; an antithrombic agent such as argatroban hydrate; and the like.
When the predetermined blood coagulation parameter is thrombo test or whole blood coagulation time, the anticoagulant drug can include an antiplatelet drug. The antiplatelet drug can include ticlopidine hydrochloride, clopidogrel sulfate, prasugrel hydrochloride, ticagrelor, a clopidogrel sulfate/aspirin combination drug, cilostazol, ethyl icosapentate, beraprost sodium, sarpogrelate hydrochloride, an aspirin/dialuminate combination drug, aspirin, and the like.
Examples of the cause derived from a blood collection technique can include: a case of contamination of heparin during blood collection such as line blood collection or post-dialysis blood collection; a case of a small blood collection amount relative to the proportion of an anticoagulant agent filled in advance in a blood collection tube; a case where the blood vessel is thin and blood collection is difficult, resulting in contamination of tissue fluid during blood collection; a case where time has elapsed from blood collection; and the like.
The cause of the prolongation of the blood coagulation time outputted from the output layer 60b of the deep learning algorithm 60 is, preferably, at least one selected from the group consisting of a cause related to prolongation of activated partial thromboplastin time, and a cause related to prolongation of prothrombin time.
Next, the controller 201 advances to step S26 shown in
The information regarding the cause of the prolongation of the blood coagulation time can include at least a label indicating a cause candidate for the prolongation of the blood coagulation time. Preferably, said information can include a probability that the cause candidate indicated by the label is a cause of the prolongation of the blood coagulation time. The label indicating a cause candidate for the prolongation of the blood coagulation time may be character information or a label value. The character information can include a notation in the form of an abbreviation or the like such as FIX or FVIII. The label value can include a notation in the form of a symbol, such as 1, 2, or the like, associated in advance with a predetermined prolongation cause.
The probability that the cause candidate is a cause of the prolongation of the blood coagulation time may be expressed in a softmax form or a binary form. The softmax form is a form that indicates the probability that, among a plurality of cause candidates for prolongation of blood coagulation time, a predetermined cause candidate is a cause of the prolongation of the blood coagulation time. For example, when the prolongation cause outputted from the deep learning algorithm 60 includes only factor VIII deficiency and factor IX deficiency as the causes whose probability is greater than 0%, the probabilities of the respective prolongation causes are indicated such that the total of these probabilities becomes 100%. The binary form is a form in which the probability that each of a plurality of prolongation cause candidates is a cause of the prolongation of the blood coagulation time, and the total of the probabilities of the prolongation cause candidates does not necessarily become 100%. The probability that a cause candidate is a cause of the prolongation of the blood coagulation time may be outputted in a form of a graph as shown, for example, in
When the information regarding the cause of the prolongation of the blood coagulation time is to be outputted, a label indicating a cause candidate, for the prolongation of the blood coagulation time, for which the probability is the highest may be outputted, for example. In this case, for example, in the examples of
When the information regarding the cause of the prolongation of the blood coagulation time is to be outputted, a label indicating a cause candidate, for the prolongation of the blood coagulation time, for which the probability is not less than a predetermined threshold may be outputted, for example. The threshold can be 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90%, for example. In the examples of
In a case where the information regarding the cause of the prolongation of the blood coagulation time is to be outputted, when a threshold is to be used, the controller 201 can retrieve the threshold from the reference value/threshold DB 202d stored in the storage 202. The threshold may be inputted by the operator through the display 4, and the controller 201 may receive the threshold.
In step S27 shown in
For example, among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is liver disease is the highest, the additional test can be a biochemical test (AST, ALT, γ-GTP, LDH, ALP, etc.) for evaluating the liver function.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is disseminated intravascular coagulation is the highest, the additional test can be platelet number measurement, blood fibrinogen concentration measurement, FDP/D-dimer measurement, blood thrombin-antithrombin complex concentration measurement, measurement of a fibrinolytic system marker, such as plasmin-α2 plasmin inhibitor complex, or the like.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is vitamin K deficiency is the highest, the additional test can be blood vitamin K concentration measurement.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is hemorrhaging is the highest, the additional test can be red blood cell number measurement, hemoglobin concentration measurement, hematocrit value measurement, platelet number measurement, or the like.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is decrease, deficiency, or dysfunction of a certain coagulation factor is the highest, the additional test can be: measurement (including a blood coagulation method, a synthetic substrate method, etc.) of the activity of the certain coagulation factor; measurement of the concentration of protein of the certain coagulation factor by an enzyme immunoassay, etc.; or the like.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is the presence of a certain coagulation factor inhibitor is the highest, the additional test can be: detection of the certain coagulation factor inhibitor, for example, detection of an anti-coagulation factor antibody by an enzyme immunoassay, etc.; a mixing test; a cross-mixing test; or the like.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is the presence of lupus anticoagulant is the highest, the additional test can be: detection of an anti-phospholipid antibody by an enzyme immunoassay, etc.; a cross-mixing test; or the like.
When the deep learning algorithm 60 has outputted that the cause of the prolongation of the blood coagulation time is the presence of an abnormal protein, the additional test can be serum protein analysis.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is an anticoagulant drug is the highest, the additional test can be confirmation of the medication history of the subject from whom the blood sample has been collected.
Among the causes of the prolongation of the blood coagulation time outputted by the deep learning algorithm 60, when the probability that the cause is a cause derived from the blood collection technique is the highest, the additional test can be a re-test including re-collection of a blood sample.
The information regarding the additional test may be a label value or character information indicating the name of the additional test.
In the output of information regarding the additional test, in addition to the additional test that corresponds to the prolongation cause having the highest probability, an additional test that corresponds to the prolongation cause having the second highest probability or lower may be outputted. In this case, a priority ranking of each additional test may be outputted.
It should be noted that step S27 can be omitted.
Further, when a more specific cause of the prolongation of the blood coagulation time has been identified by the additional test, the operator may input, through the display 4, the identified prolongation cause in association with identification information of the blood sample for which the additional test has been performed. The inputted prolongation cause may be stored into the additional test DB 202e of the storage 202, so as to be associated with the identification information of the blood sample and the data group including the plurality of data with respect to the blood sample.
An embodiment disclosed in the present specification relates to a training apparatus 5 (hereinafter, simply referred to as a “training apparatus 5”) of the deep learning algorithm 50.
The controller 501 includes an arithmetic processing device such as a CPU.
The storage 502 stores: a training program 502b described later; an algorithm database (DB) 502c storing one or a plurality of deep learning algorithms 50 and/or deep learning algorithms 60; and a training data database (DB) 502d storing the first training data and the second training data so as to be associated with each other.
The input interface 506 receives input information inputted by the operator through the input unit 511 implemented as a touch panel, a keyboard, or the like, and transmits the received input information to the controller 501 or the storage 502.
The output interface 507 transmits output information outputted by the controller 501, to the output unit 512 such as a display.
The communication interface 508 communicably connects the training apparatus 5 to a network 95. The connection may be wired connection or wireless connection.
The media interface 509 performs transmission of information with respect to a nonvolatile storage medium such as a CD-ROM, a DVD-ROM, an external hard disk, or the like.
Signal transmission in the training apparatus 5 is performed via the bus 510.
In step S51, the controller 501 receives a process start request inputted by the operator through the input unit 511, and retrieves training data from the training data DB 502d.
Subsequently, in step S52, the controller 501 retrieves a deep learning algorithm 50 or a deep learning algorithm 60 from the algorithm DB 502c in the storage 502, inputs the training data to the retrieved deep learning algorithm, and performs training. Details of the training have been described in “i. Training of deep learning algorithm”. Inputting training data to the deep learning algorithm 60 to train the deep learning algorithm 60 is also referred to as re-training of the deep learning algorithm 60.
Next, in step S53, the controller 501 determines whether or not the deep learning algorithm 50 or the deep learning algorithm 60 has been trained by using all of the training data that should be used. When training has been performed by using all of the training data (“YES” in step S53), the controller 501 advances to step S54, and stores the trained deep learning algorithm 60 into the algorithm DB 502c stored in the storage 202.
In step S53, when training has not been performed by using all of the training data (“NO” in step S53), the controller 501 returns to step S51 and continues the training process.
The analysis program 202b and/or the training program 502b can be provided as a program product such as a storage medium. The computer program is stored in a storage medium such as a hard disk, a semiconductor memory device such as a flash memory, or an optical disk. The storage form of the program in the storage medium is not limited as long as the controller can read the program.
It should be noted that, in the present specification, the anticoagulant agent and the anticoagulant drug are used so as to be distinguished from each other. The anticoagulant agent is an agent that is filled in a blood collection tube or a syringe in order to prevent deposition of fibrin during blood collection. For example, usually, when blood is collected to be used in a coagulation test, the collection is preferably performed by using, as the anticoagulant agent, a citrate, e.g., a sodium citrate solution. The blood specimen is prepared by, for example, using a 3.1% to 3.3% (weight/volume) trisodium citrate solution as an anticoagulant agent and mixing this anticoagulant agent and whole blood at a volume ratio of about 1:8.5 to 1:9.5. Alternatively, the blood specimen may be plasma separated from a mixture of the anticoagulant agent and whole blood.
As the known blood analyzer 100, a fully automated blood coagulation measurement apparatus CN-6000, CN-3000, or the like manufactured by Sysmex Corporation can be used, for example.
5-1. Comparison with Conventional Method
APTT measurement was performed by using specimens (plasma) for which the APTT prolongation cause is the presence of lupus anticoagulant (LA), and specimens for which the APTT prolongation cause is the presence of a factor VIII inhibitor (FVIII inhibitor). Using the measurement result, the prediction accuracy of a conventional method (first-order differential method) and the prediction accuracy of the present analysis method executed by the analyzer 1 were compared with each other.
Number | Date | Country | Kind |
---|---|---|---|
2020-171206 | Oct 2020 | JP | national |