Aggravation prediction device, aggravation prediction method, and program

The severity prediction system uses blood test values and logistic regression to predict disease severity in virus-infected individuals, addressing the challenge of burdensome medical observations and improving prediction accuracy.

JP2024157060AInactive Publication Date: 2024-11-07KEIO UNIV
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2021123710
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2021-07-28
Publication Date
2024-11-07
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing methods cannot accurately predict whether a virus-infected individual will become seriously ill, requiring extensive medical observation and testing, which burdens healthcare personnel.

Method used

A severity prediction system that extracts feature amounts from blood test values of infected individuals using logistic regression analysis to construct a predictive model, allowing for simpler determination of disease severity.

Benefits of technology

Enables accurate prediction of disease severity in virus-infected individuals using a simpler method, reducing the burden on medical personnel and enabling timely preparation for severe cases.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024157060000001_ABST
    Figure 2024157060000001_ABST
Patent Text Reader

Abstract

To perform prediction on aggravation of a virus-infected person with a simpler method.SOLUTION: An aggravation prediction device 20 comprises a feature quantity extraction unit 212, a prediction model construction unit 213, and a prediction execution unit 214. The feature quantity extraction unit 212 extracts, as feature quantities, inspection values of a plurality of components included in the blood of a virus-infected person. The prediction model construction unit 213 performs logistic regression analysis, with the feature quantities extracted from the blood of the infected person to be a subject as a factor, and with whether the infected person to be the subject is aggravated due to infection with the virus as an objective variable, and thereby constructs a prediction model for performing prediction on the aggravation. The prediction execution unit 214 performs the prediction on the aggravation of the infected person to be a prediction target person on the basis of the feature quantities extracted from the blood of the infected person to be the prediction target person and the prediction model.SELECTED DRAWING: Figure 2
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an aggravation prediction device, an aggravation prediction method, and a program. [Background technology]

[0002] In recent years, measures against various viral infections, such as the new coronavirus, have become more important. Accordingly, the development of technologies related to viral infection has become widespread.

[0003] An example of such a technology is disclosed in Patent Document 1. The technology disclosed in Patent Document 1 improves the accuracy of negative determination by avoiding situations in which false negatives occur in so-called PCR testing based on the polymerase chain reaction (PCR). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2019-092505 A Summary of the Invention [Problem to be solved by the invention]

[0005] However, if a patient infected with a virus becomes seriously ill, treatment becomes more difficult because specialized medical staff, special medical equipment, drugs, etc. are required. Therefore, in order to prepare for such a worsening of the condition, it is necessary in the medical field to predict whether a patient will become seriously ill or not.

[0006] In this regard, general technologies such as the technology disclosed in the above-mentioned Patent Document 1 can only determine whether or not a person is infected with a virus (i.e., whether the result is positive or negative), but cannot determine whether or not a person infected with the virus will subsequently develop severe symptoms.

[0007] For this reason, medical professionals observe infected individuals as needed, conduct tests on many items, and predict whether the disease will worsen based on various indicators (e.g., level of consciousness, fever, cough, taste disorder, fatigue, nausea and vomiting, heart failure, and liver dysfunction, etc.). However, predicting whether the disease will worsen, which requires such timely observation and testing on many items, places a burden on medical professionals.

[0008] The present invention has been made in view of the above circumstances. An object of the present invention is to predict the progression of a virus-infected person's condition using a simpler method. [Means for solving the problem]

[0009] In order to solve the above problems, a system for predicting aggravation according to one embodiment of the present invention comprises: A feature extraction means for extracting test values ​​of each of a plurality of components contained in the blood of a person infected with a virus as features; a model construction means for constructing a prediction model for predicting the onset of severity by performing a logistic regression analysis using the feature amount extracted from the blood of the subject infected person as a factor and using whether the subject infected person has become severe due to infection with the virus as a response variable; A prediction means for predicting the severity of the infected person who is a prediction target based on the feature amount extracted from the blood of the infected person who is a prediction target and the prediction model; The present invention is characterized by comprising: Effect of the Invention

[0010] According to the present invention, it is possible to predict the onset of symptoms in a virus-infected person using a simpler method. [Brief description of the drawings]

[0011] [Figure 1] FIG. 2 is a block diagram showing the overall configuration of the severity prediction system S. [Diagram 2]1 is a schematic diagram showing an overview of the processing performed by the aggravation prediction device 20 to realize prediction of the aggravation of a virus-infected person. FIG. [Diagram 3] 1 is a block diagram showing an example of a configuration of a test data management device 10. FIG. [Figure 4] 4 is a table showing a schematic structure of test data stored in test data storage unit 151. [Diagram 5] FIG. 2 is a block diagram showing an example of the configuration of the aggravation prediction device 20. [Figure 6] 13 is a flowchart illustrating the flow of a test data update process executed by the test data management device 10 and the severity prediction device 20. [Figure 7] 13 is a flowchart illustrating the flow of a prediction model construction process executed by the severity prediction device 20. [Figure 8] 13 is a flowchart illustrating the flow of a prediction execution process executed by the severity prediction device 20. [Figure 9] 13 is a graph showing the change in the degree of influence of each factor depending on the size of the training group. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

[0013] [System Configuration] Fig. 1 is a block diagram showing the overall configuration of an aggravation prediction system S according to this embodiment. As shown in Fig. 1, the aggravation prediction system S includes n (n is any integer value equal to or greater than 1) test data management devices 10 (here, test data management devices 10-1, ..., test data management devices 10-n), a severity prediction device 20, and a network 30. In the following description, when the n test data management devices 10 are described without distinction, the suffixes of the reference numerals are omitted and the devices are simply referred to as "test data management devices 10."

[0014] Each of the test data management devices 10 and the severity prediction device 20 are communicatively connected in accordance with an arbitrary communication method. This communication may be performed directly between the devices, or may be performed via a network 30 including a relay device. When communication is performed via the network 30, the network 30 is realized by a network such as the Internet or a LAN (Local Area Network).

[0015] The severe illness prediction system S is an example of an embodiment of the present invention, and predicts the severe illness of a virus-infected person in a simpler manner. Here, the inventor of the present invention conducted repeated testing and research on the prediction of the severe illness of a virus-infected person, and found that there is a correlation between the test values ​​of multiple components contained in the blood of a virus-infected person and the severe illness of the virus-infected person. The inventor of the present invention then came up with the idea that it is possible to predict the severe illness of a virus-infected person based on the test values ​​of a specific component contained in the blood of the virus-infected person, and thus came up with the present invention.

[0016] The following is a detailed description of an example of an embodiment of the present invention, that is, the severe illness prediction system S. As an example for the purposes of this description, it is assumed that the "virus" is SARS-CoV-2, also known as the new coronavirus, and that an "infected person" of the virus is a person who has been infected with the new coronavirus (COVID-19) by this new coronavirus. However, this is merely an example for the purpose of explanation and is not intended to limit the scope of application of the present invention. For example, the severe illness prediction system S may be configured to predict the severe illness of a person infected with another virus.

[0017] The test data management device 10 manages test data. The test data management device 10 is realized by, for example, a server or a personal computer installed in a medical facility such as a hospital or a testing institute, or a cloud server. The test data includes, for example, test values ​​of each of a plurality of components contained in the blood of a person infected with a virus, and information indicating whether the person infected with the virus has become seriously ill. When a blood test is performed at a medical facility or when a medical professional diagnoses that a person infected with a virus has become seriously ill, the test data management device 10 appropriately updates the test data according to the test or diagnosis results. In addition, the test data management device 10 provides the latest updated test data to the severe illness prediction device 20.

[0018] The aggravation prediction device 20 predicts the aggravation of a virus-infected person based on the test data provided by the test data management device 10. The aggravation prediction device 20 is realized by, for example, a server or a personal computer installed in a medical facility such as a hospital or a testing institute, or a cloud server.

[0019] An overview of the process performed by the severity prediction device 20 will be described with reference to Fig. 2. Fig. 2 is a schematic diagram showing an overview of the process performed by the severity prediction device 20 to realize prediction regarding the severity of a virus-infected person.

[0020] As a premise, in this embodiment, people infected with a virus are classified into three types. First, individuals infected with the virus whose test data will be used to build a prediction model are classified as "subjects." These subjects are further classified into those diagnosed with severe disease (i.e., those who developed severe disease) (corresponding to (A) in the figure) and those diagnosed with non-severe disease (i.e., those who did not develop severe disease) (corresponding to (B) in the figure). On the other hand, individuals infected with the virus who are the targets of predictions regarding progression are classified as "prediction targets" (corresponding to (C) in the figure).

[0021] Then, the severe illness prediction device 20 extracts, as features, test values ​​of each of a plurality of components contained in the blood of the subject infected with the virus (corresponding to (1) and (2) in the figure). Next, the severe illness prediction device 20 uses the feature amounts extracted from the blood of the infected subject as factors (i.e., explanatory variables) and whether the infected subject from which the feature amounts were obtained has become seriously ill due to infection with the virus as the objective variable to perform logistic regression analysis to construct a prediction model that predicts the severity of illness (corresponding to (3) in the figure). This prediction model is a binary classifier that classifies the predicted subject into either "will become seriously ill" or "will not become seriously ill."

[0022] Next, the severe illness prediction device 20 extracts, as features, test values ​​of each of a plurality of components contained in the blood of the person infected with the virus that is to be predicted (corresponding to (4) in the figure). Furthermore, the severe illness prediction device 20 predicts the severity of the infected person to be predicted based on the feature amount extracted from the blood of the infected person to be predicted and the prediction model. Specifically, the severe illness prediction device 20 inputs the feature amount extracted from the infected person to be predicted into the prediction model (corresponding to (5) in the figure). Then, the severe illness prediction device 20 presents the output from the prediction model corresponding to this input to the user of the severe illness prediction device 20 (e.g., a medical worker) as a prediction result regarding the severity of the predicted person (corresponding to (6) in the figure).

[0023] In this way, the severe illness prediction device 20 predicts the severe illness of the infected person to be predicted based on the prediction model constructed from the feature values ​​extracted from the blood of the subject infected person and the feature values ​​extracted from the blood of the infected person to be predicted. Here, the feature values ​​for realizing such processing can be extracted from the test values ​​of each of the multiple components contained in the blood of the infected person. That is, medical personnel can make predictions by simply taking blood samples from infected individuals using the severity prediction device 20. This is a much simpler method than the conventional predictions of severity that require appropriate observations and tests for many items. Therefore, the severe illness prediction device 20 can predict the severe illness of a virus-infected person in a simpler manner.

[0024] Next, the hardware and functional configurations of the test data management device 10 and the severity prediction device 20 for realizing prediction of the severity of such virus-infected individuals will be described.

[0025] [Configuration of the test data management device] The configuration of the test data management device 10 will be described with reference to Fig. 3. Fig. 3 is a block diagram showing an example of the configuration of the test data management device 10. 3, the test data management device 10 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a communication unit 14, a storage unit 15, an input unit 16, an output unit 17, and a drive 18. These units are connected by signal lines and transmit and receive signals between each other.

[0026] The CPU 11 executes various processes according to a program recorded in the ROM 12 or a program loaded from the storage unit 15 to the RAM 13 . The RAM 13 also stores data and the like necessary for the CPU 11 to execute various processes.

[0027] The communication unit 14 performs communication control for the CPU 11 to communicate with other devices. The storage unit 15 is configured with a semiconductor memory such as a dynamic random access memory (DRAM), and stores various data.

[0028] The input unit 16 is composed of various buttons and a touch panel, or an external input device such as a mouse and a keyboard, and inputs various information in response to user instructions. The output unit 17 is composed of a display, a speaker, etc., and outputs images and sounds.

[0029] Removable media 51, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is appropriately loaded into the drive 18. A program read from the removable media 51 by the drive 18 is installed in the storage unit 15 as necessary.

[0030] In the test data management device 10, these units cooperate with each other to perform the "test data management process." Here, the test data management process is a series of processes in which the test data management device 10 appropriately updates the management data and provides the latest updated test data to the severity prediction device 20.

[0031] When the test data management process is executed, as shown in FIG. 3, in the CPU 11, a test data management unit 111 and a test data providing unit 112 function. In addition, in one area of ​​the storage unit 15, an inspection data storage unit 151 is provided. Including cases not specifically mentioned below, data required to realize processing is transmitted and received between these functional blocks at appropriate times as appropriate.

[0032] The test data management unit 111 acquires the latest test data (e.g., the latest test values ​​and the latest diagnostic results) based on input operations by a user of the test data management device 10 accepted by the input unit 16 and test results from other devices (e.g., blood test devices used in medical facilities) received via the communication unit 14. Then, the test data management unit 111 updates the existing test data stored in the test data storage unit 151 to the latest content based on the acquired latest test data. That is, the test data storage unit 151 functions as a storage unit that stores the test data.

[0033] The test data stored in the test data storage unit 151 will be described with reference to Fig. 4. Fig. 4 is a table showing a schematic structure of the test data stored in the test data storage unit 151. As shown in Fig. 4, this test data table includes columns of infected person information such as "infected person ID" and "infected person attributes," and test values ​​such as "platelets," "LDH," ..., "ferritin," and "severe disease diagnosis result." This test data table also has a row (record) for each infected person. Corresponding data is stored in each input item (field) by the test data management unit 111.

[0034] Here, the infected person ID is an identifier for identifying each infected person, and a unique number, text, or other information is assigned to each infected person and stored.

[0035] The infected person attributes are information indicating the attributes of an infected person, and for example, information such as the infected person's age, gender, and whether or not they have an underlying disease is stored.

[0036] Platelets, LDH,..., and ferritin correspond to the respective components of the infected person's blood, and the test values ​​for each of these blood components are stored. However, these components (i.e., platelets, LDH,..., and ferritin) shown in the figure are merely examples, and test values ​​for components other than these may be stored. Note that the methods of blood collection and blood testing, which are the basis for obtaining test values, are well known to those skilled in the art, and therefore a detailed explanation will be omitted here. Also, in the figure, the test values ​​are indicated as "***".

[0037] The severity diagnosis result stores the diagnosis result of the severity by the medical staff. Specifically, information such as "Severe" is stored if the condition is diagnosed as severe, "Non-severe" is stored if the condition is not diagnosed as severe, and "Unconfirmed" is stored if the diagnosis has not been made yet.

[0038] As described above, when blood tests are conducted at a medical facility or when a medical professional diagnoses that a person infected with a virus has become seriously ill, the test data management unit 111 appropriately updates the test data in accordance with the results of these tests and diagnoses. The test data may be test data collected specifically for this embodiment, or may be general test data collected for the purpose of sharing information between medical facilities or for research. In other words, this embodiment can use a large amount of general test data that already exists in the world to build a prediction model, etc.

[0039] Returning to Fig. 3, the test data providing unit 112 provides the latest updated test data stored in the test data storage unit 151 by transmitting it to the severity prediction device 20 via the communication unit 14. The timing of the provision may be every time the test data is updated, in response to a request from the severity prediction device 20, or at a predetermined cycle. In addition, the method of provision may be a method of transmitting the test data to the severity prediction device 20 via the communication unit 14, or a method via a storage medium such as the removable media 51.

[0040] [Configuration of the severity prediction device] Next, the configuration of the aggravation prediction device 20 will be described with reference to FIG. 5. FIG. 5 is a block diagram showing an example of the configuration of the aggravation prediction device 20. As shown in FIG. 5, the aggravation prediction device 20 includes a CPU 21, a ROM 22, a RAM 23, a communication unit 24, a storage unit 25, an input unit 26, an output unit 27, and a drive 28. These units are connected by signal lines and transmit and receive signals to each other. These units are realized by hardware equivalent to the units of the same name included in the test data management device 10 described above with reference to FIG. 3. For example, the CPU 21 is realized by hardware equivalent to the CPU 11. Therefore, here, overlapping descriptions of these units will be omitted.

[0041] In the severity prediction device 20, these parts work together to perform "test data update processing," "prediction model construction processing," and "prediction execution processing." As described above, the test data management process is a series of processes in which the test data management device 10 appropriately updates the management data and provides the latest updated test data to the severity prediction device 20.

[0042] The prediction model construction process is a series of processes for constructing a prediction model for predicting the progression of disease based on features extracted from the test data of the subject.

[0043] Furthermore, the prediction execution process is a series of processes that executes predictions regarding progression based on features extracted from the test data of the prediction subject and a prediction model.

[0044] When these processes are executed, as shown in FIG. 5, in the CPU 21, an inspection data acquisition unit 211, a feature amount extraction unit 212, a prediction model construction unit 213, and a prediction execution unit 214 function. In addition, in one area of ​​the storage unit 25, an inspection data storage unit 251, a feature amount storage unit 252, and a prediction model storage unit 253 are provided. Including cases not specifically mentioned below, data required to realize processing is transmitted and received between these functional blocks at appropriate times as appropriate.

[0045] The test data acquiring unit 211 acquires the latest test data provided from the test data managing device 10. For example, when the test data is provided by transmission from the test data managing device 10, the test data acquiring unit 211 acquires the test data by receiving it via the communication unit 24. When the test data is provided via a storage medium such as the removable medium 51, the test data acquiring unit 211 acquires the test data by reading it via the drive 28.

[0046] Then, the test data acquiring unit 211 stores the acquired test data in the test data storage unit 251. That is, the test data storage unit 251 functions as a storage unit that stores test data. As a result, the test data stored in the test data storage unit 251 is updated to the latest content, similar to the test data stored in the test data storage unit 151. Note that the test data may be acquired from any of the multiple test data management devices 10.

[0047] Feature extraction unit 212 extracts features from the test data stored in test data storage unit 251. Specifically, feature extraction unit 212 extracts, as a feature, information that pairs the test values ​​of blood components of a virus-infected person with the diagnosis result of aggravation of the infected person for each infected person. Here, the test data includes test values ​​of various components in blood, among which there are components that are highly correlated with the worsening of the disease in a virus-infected person and components that are less correlated (or have no correlation). Therefore, the feature extraction unit 212 does not extract the test values ​​of all components as features, but extracts the test values ​​of components that are highly correlated with the worsening of the disease in a virus-infected person. This makes it possible to build a prediction model that can predict the worsening of the disease with higher accuracy, or to execute predictions. Specifically, which component's test value is desirable to extract as a feature will be described in the verification example described later.

[0048] Then, the feature amount extraction unit 212 stores the extracted feature amount in the feature amount storage unit 252. That is, the feature amount storage unit 252 functions as a storage unit that stores the feature amount.

[0049] The prediction model construction unit 213 constructs a prediction model based on the features of the test data of the subjects stored in the feature storage unit 252. As described above, the construction of this prediction model uses the features of the test data of the subjects who are infected and have become severe (i.e., the infected subjects whose severity diagnosis result is "severe") and the features of the test data of the subjects who are infected and have not become severe (i.e., the infected subjects whose severity diagnosis result is "non-severe").

[0050] The prediction model construction unit 213 first performs a predetermined pre-processing on these features in order to properly construct a prediction model. For example, as pre-processing, the prediction model construction unit 213 standardizes data for each test value of the blood components extracted as the features, using the average and standard deviation of the test values, so that the average is 0 and the standard deviation is 1. This makes it possible to handle each blood component on the same scale and construct a prediction model.

[0051] In addition, for example, there may be cases where some of the test values ​​of each of the multiple components extracted as features for a certain infected person are missing. For example, there may be cases where there is no test value for a certain component in the test data to begin with. In this case, as a preprocessing step, the prediction model construction unit 213 interpolates the missing values ​​using a representative value based on the test values ​​for this certain component for other infected people. The representative value may be, for example, the average value, the median value, or the mode value. This makes it possible to use the features of a subject in which test values ​​for some blood components are not missing but test values ​​exist for other blood components in constructing a prediction model.

[0052] Then, the prediction model construction unit 213 constructs a prediction model based on the feature amount that has been subjected to such preprocessing. Here, the prediction model constructed by the prediction model construction unit 213 may be a binary classifier that classifies a prediction target person into either "will become seriously ill" or "will not become seriously ill", and the construction method is not particularly limited. For example, the prediction model construction unit 213 constructs a prediction model that predicts on the onset of aggravation by performing a logistic regression analysis using the feature amount extracted from the blood of the subject infected person as a factor (i.e., explanatory variable) and whether or not the subject infected person from which the feature amount was obtained has become seriously ill due to infection with the virus as a response variable.

[0053] In this case, the prediction model construction unit 213 optimizes explanatory variables by a variable increase / decrease method in the logistic regression equation shown in the following formula (1). Here, the explanatory variables are the test values ​​of the blood components in the features extracted by the feature extraction unit 212, and a prediction model is constructed by performing machine learning to optimize which partial regression coefficients of the test values ​​of the blood components are to be increased or decreased as explanatory variables.

[0054]

number

[0055] In this case, the prediction model construction unit 213 may further apply other existing methods in constructing the prediction model. For example, the prediction model construction unit 213 adjusts class weights based on the ratio of infected subjects who have become seriously ill to infected subjects who have not become seriously ill.

[0056] Alternatively, the prediction model construction unit 213 performs validation by k-fold cross-validation in order to achieve generalization performance. In this case, the value of K is set to 5, for example.

[0057] In addition, the prediction model construction unit 213 creates a receiver operating characteristic (ROC) curve as an index of the predictive ability of the classifier, and calculates the area under the curve (AUC). c If p > p, the condition is considered non-severe (i.e., negative), and the cutoff value p c If p is less than or equal to the test result, the condition is considered severe (i.e., positive), and based on these results, the sensitivity (i.e., the probability that a normal person can be determined to be normal) and specificity (i.e., the probability that an abnormal person can be determined to be above or equal to the test result) are determined and verified.

[0058] The prediction model construction unit 213 repeats machine learning while appropriately performing such verification, etc., to construct a prediction model. Then, the prediction model construction unit 213 stores the constructed prediction model in the prediction model storage unit 253. That is, the prediction model storage unit 253 functions as a storage unit that stores the prediction model.

[0059] The prediction execution unit 214 predicts the progression of the disease based on the features of the test data of the infected person to be predicted, which are stored in the feature storage unit 252, and the prediction model stored in the prediction model storage unit 253. For example, the prediction execution unit 214 acquires the features of the infected person to be predicted, based on a selection operation of the infected person to be predicted from the user received by the input unit 26 (or a selection operation of the infected person to be predicted from the user received via the communication unit 24). In this case, the infected person to be predicted is selected from the infected people whose severe progression diagnosis result is "undetermined".

[0060] In addition, in order to perform an appropriate prediction, the prediction execution unit 214 performs preprocessing such as data normalization on the feature amount of the prediction target person in the same manner as the preprocessing by the prediction model construction unit 213. The prediction execution unit 214 then executes a prediction regarding the worsening of the condition by inputting the feature amount of the prediction target person after this preprocessing into the prediction model stored in the prediction model storage unit 253. The prediction execution unit 214 also sets the output from the prediction model corresponding to this input as the prediction result regarding the worsening of the condition of the prediction target person. Here, this prediction model is a binary classifier that classifies the prediction target person into either "will become seriously ill" or "will not become seriously ill". Therefore, the output of the prediction model indicates whether the prediction target person corresponds to "will become seriously ill" or "will not become seriously ill".

[0061] Then, the prediction execution unit 214 presents this prediction result to a user (e.g., a medical professional) of the severity prediction device 20. This presentation may be, for example, a display on a display included in the output unit 27, an audio output from a speaker included in the output unit 27, printing on a paper medium from a printing device (not shown) via the communication unit 24, or transmission to another device (not shown) used by the user via the communication unit 24.

[0062] This makes it possible to predict the severity of the infected person who is the prediction target based on the prediction model constructed from the feature values ​​extracted from the blood of the subject infected person and the feature values ​​extracted from the blood of the infected person who is the prediction target, and present the prediction result to the user. Therefore, the user can take measures (for example, securing hospital beds and special medical equipment, etc.) in advance for the infected person who is predicted to become seriously ill, and start appropriate treatment as soon as possible.

[0063] [Operation] Next, the operation of the severity prediction system S will be described. [Inspection data update process] 6 is a flowchart explaining the flow of the test data update process executed by the test data management device 10 and the aggravation prediction device 20. The test data update process is periodically and repeatedly executed in response to the start-up of the test data management device 10 and the aggravation prediction device 20.

[0064] First, the flow of processing on the test data management device 10 side will be described. In step S11, test data management unit 111 determines whether the latest test data has been acquired by a user's input operation or by receiving test results from another device (for example, a blood test device used in a medical facility). If the latest test data has been acquired, step S11 is determined as Yes, and the process proceeds to step S12. On the other hand, if the latest test data has not been acquired, step S11 is determined as No, and the process proceeds to step S13.

[0065] In step S12, the test data management unit 111 stores the acquired latest test data in the test data storage unit 151, thereby updating the test data.

[0066] In step S13, the test data providing unit 112 judges whether it is time to provide the latest updated test data stored in the test data storage unit 151 by transmitting the data to the severity prediction device 20 via the communication unit 14. As described above, the timing of providing the data may be each time the test data is updated, in response to a request from the severity prediction device 20, or at a predetermined cycle. If it is time to provide the test data, the result of step S13 is judged as Yes, and the process proceeds to step S14. On the other hand, if it is not time to provide the test data, the result of step S13 is judged as No, and the process ends. Then, the process is repeated from step S11 at a predetermined cycle.

[0067] In step S14, the test data providing unit 112 provides the latest updated test data stored in the test data storage unit 151 by transmitting it to the severity prediction device 20 via the communication unit 14. This ends the process. Then, the process is repeated from step S11 at a predetermined cycle.

[0068] Next, the flow of processing on the side of the severity prediction device 20 will be described. In step S21, the test data acquisition unit 211 determines whether the latest test data has been received. If the latest test data has been received, the result of step S21 is Yes, and the process proceeds to step S22. On the other hand, if the latest test data has not been received, the result of step S21 is No, and the process ends. Then, the process is repeated from step S21 at a predetermined cycle.

[0069] In step S22, the inspection data is updated by storing the latest received inspection data in the inspection data storage unit 251. This ends the process. Then, the process is repeated from step S21 at a predetermined cycle.

[0070] By the above-described test data update process, the test data stored in the test data storage unit 151 and the test data storage unit 251 is appropriately updated to the latest content.

[0071] [Prediction model construction process] 7 is a flowchart illustrating the flow of a prediction model construction process executed by the severity prediction device 20. The prediction model construction process is executed based on a start instruction operation of the prediction model construction process by the user of the severity prediction device 20, which is received by the input unit 26.

[0072] In step S31, feature extraction section 212 extracts features of the subject from the test data stored in test data storage section 251.

[0073] In step S32, feature extraction section 212 stores the extracted features of the subject in feature storage section 252.

[0074] In step S33, the prediction model construction unit 213 performs a predetermined pre-processing on the features of the subject in order to appropriately construct a prediction model.

[0075] In step S34, the prediction model construction unit 213 constructs a prediction model based on the pre-processed feature amounts of the subject.

[0076] In step S35, the prediction model construction unit 213 judges whether or not to end the construction of the prediction model. For example, the prediction model construction unit 213 ends the construction of the prediction model when machine learning for constructing the prediction model has been continued a predetermined number of times or a predetermined time, or when the prediction accuracy of the prediction model has reached a predetermined accuracy. If the construction of the prediction model is to be ended, it is judged as Yes in step S35, and the process proceeds to step S36. On the other hand, if the construction of the prediction model is not to be ended, it is judged as No in step S35, and the process returns to step S34, and the construction of the prediction model is continued.

[0077] In step S36, the prediction model construction unit 213 stores the constructed prediction model in the prediction model storage unit 253. This ends the process.

[0078] By the prediction model construction process described above, a prediction model for predicting the worsening of symptoms is constructed.

[0079] [Prediction execution process] 8 is a flowchart illustrating the flow of a prediction execution process executed by the severity prediction device 20. The prediction execution process is executed based on a start instruction operation of the prediction execution process by the user of the severity prediction device 20, which is received by the input unit 26.

[0080] In step S41, the feature amount of the prediction target is extracted from the test data stored in the test data storage unit 251.

[0081] In step S42, the feature extraction unit 212 stores the extracted features of the prediction target person in the feature storage unit 252.

[0082] In step S43, the prediction execution unit 214 performs pre-processing such as data normalization on the feature amounts of the person to be predicted in the same manner as the pre-processing performed by the prediction model construction unit 213 in order to perform appropriate prediction.

[0083] In step S44, the prediction execution unit 214 executes a prediction regarding the worsening of the condition by inputting the feature amount of the prediction subject after preprocessing into the prediction model stored in the prediction model storage unit 253.

[0084] In step S45, the prediction execution unit 214 presents the output from the prediction model corresponding to this input as a prediction result regarding the severity of the prediction subject to the user (e.g., medical staff) of the severity prediction device 20. This ends the present process.

[0085] Through the prediction execution process described above, the severe illness prediction device 20 predicts the severe illness of the infected person to be predicted based on the prediction model constructed from the features extracted from the blood of the subject infected person and the features extracted from the blood of the infected person to be predicted. Here, the features for realizing such a process can be extracted from the test values ​​of each of the multiple components contained in the blood of the infected person. That is, medical personnel can make predictions by simply taking blood samples from infected individuals using the severity prediction device 20. This is a much simpler method than the conventional predictions of severity that require appropriate observations and tests for many items. Therefore, the severe illness prediction device 20 can predict the severe illness of a virus-infected person in a simpler manner.

[0086] [Verification example] The embodiment of the present invention has been described above. Next, a verification example in the embodiment of the present invention will be described. In this verification example, the above-mentioned processes in the embodiment of the present invention were performed on the test data of 300 virus-infected individuals collected from multiple hospitals. The age groups of these 300 individuals were diverse. The genders of these 300 individuals included almost equal numbers of men and women. Furthermore, of these 300 individuals, approximately 100 were seriously ill patients.

[0087] In order to improve the accuracy of the constructed prediction model, multiple training groups were created by changing the size of the training group used to construct the prediction model (i.e., how many of the features of the 300 people were used for machine learning to construct the prediction model). Specifically, a first training group was created in which 10% of the features of the 300 people, or 30 people's features, were used to construct the prediction model, a second training group was created in which 20% of the features of the 300 people, or 60 people's features, were used to construct the prediction model, and by increasing this in 10% intervals, up to the ninth training group was created in which 90% of the features of the 270 people, or 20% of the features, were used to construct the prediction model.

[0088] Then, we ran the process of building a predictive model using the features of each size group 20 times. Furthermore, the coefficients (i.e., Z scores) of each factor (i.e., the test values ​​of each blood component that was used as an explanatory variable) obtained in constructing each prediction model were normalized. The average value of the results of 20 runs was then calculated. The average value is shown in Figure 9.

[0089] Figure 9 is a graph showing the change in the influence of each factor depending on the size of the training group. In this graph, the vertical axis shows the average coefficient (i.e., Z score) of each factor, and the horizontal axis shows the size of the training group.

[0090] As mentioned above, instead of extracting test values ​​for all components as features, test values ​​for components that are highly correlated with the worsening of symptoms in virus-infected individuals are extracted. Here, test values ​​for seven components, namely lymphocytes, hemoglobin, creatine (Cre), lactate dehydrogenase (LDH), ferritin, D-dimer, and C-reactive protein (CRP), are extracted from over 30 blood components. These test values ​​were selected from the medical perspective of doctors and from attempts to build a predictive model as components that are highly correlated with the worsening of symptoms in virus-infected individuals.

[0091] 9, in this verification example, the components highly correlated with the severity of the virus infection were lymphocytes, ferritin, hemoglobin, C-reactive protein, and lactate dehydrogenase, in that order. Therefore, when constructing a prediction model, it is desirable to construct a prediction model with the test values ​​of some or all of these components as factors (i.e., objective variables) as features, so that the factors are more important than other components contained in the blood.

[0092] In this verification example, the prediction model of each training group had a prediction accuracy of over 80% regarding the worsening of symptoms. In other words, it was possible to predict with a high probability of over 80% whether the predicted subject would become seriously ill (or not). Thus, according to the embodiment of the present invention, not only can predictions regarding the worsening of symptoms in virus-infected individuals be made in a simpler manner, but also highly accurate predictions can be made.

[0093] [Variations] Although the embodiment of the present invention has been described above, this embodiment is merely an example and does not limit the technical scope of the present invention. The present invention can take various other embodiments and can be modified in various ways, such as omissions and substitutions, without departing from the gist of the present invention. In this case, these embodiments and their modifications are included in the scope and gist of the invention described in this specification, etc., and are included in the scope of the invention described in the claims and their equivalents. As an example, the embodiment of the present invention described above may be modified as illustrated below.

[0094] The device configuration of the aggravation prediction system S in the above-mentioned embodiment is merely an example, and can be changed as appropriate. For example, in the above-mentioned embodiment, the test data management device 10 and the aggravation prediction device 20 are realized as separate devices. However, this is not limited thereto, and for example, the test data management device 10 and the aggravation prediction device 20 may be realized as an integrated device.

[0095] In the above-described embodiment, the test data management device 10 and the aggravation prediction device 20 are each realized by a single computer. However, the present invention is not limited to this, and each of the test data management device 10 and the aggravation prediction device 20 may be realized by multiple computers by using a technology such as cloud computing.

[0096] In addition, in the above-mentioned embodiment, it is assumed that one prediction model is constructed. However, it is not limited to this, and multiple prediction models may be constructed. For example, the test data is classified by gender or age group based on the attributes of the infected person contained in the test data. Then, a prediction model may be constructed for each gender and age group. Then, a prediction regarding the worsening of the condition may be made using a prediction model that is suitable for the gender and age group of the person to be predicted. This makes it possible to make a prediction that is more suitable for characteristics such as gender and age group.

[0097] In the above-described embodiment, the classifier is created using a logistic regression formula. However, the present invention is not limited to this, and the classifier may be created using other methods such as a simple perceptron or a linear support vector machine.

[0098] [Configuration example] As described above, the severe illness prediction device 20 according to this embodiment includes the feature extraction unit 212, the prediction model construction unit 213, and the prediction execution unit 214. The feature extraction unit 212 extracts, as features, test values ​​for each of a plurality of components contained in the blood of a virus-infected person. The predictive model construction unit 213 constructs a predictive model that predicts the worsening of symptoms by performing logistic regression analysis using features extracted from the blood of the infected subject as factors and whether the infected subject has become seriously ill due to infection with the virus as the objective variable. The prediction execution unit 214 predicts the progression of the disease in an infected person who is a prediction target, based on the feature amount extracted from the blood of the infected person who is a prediction target, and a prediction model.

[0099] In this way, the severe illness prediction device 20 predicts the severe illness of the infected person to be predicted based on the prediction model constructed from the feature values ​​extracted from the blood of the subject infected person and the feature values ​​extracted from the blood of the infected person to be predicted. Here, the feature values ​​for realizing such processing can be extracted from the test values ​​of each of the multiple components contained in the blood of the infected person. That is, medical personnel can make predictions by simply taking blood samples from infected individuals using the severity prediction device 20. This is a much simpler method than the conventional predictions of severity that require appropriate observations and tests for many items. Therefore, the severe illness prediction device 20 can predict the severe illness of a virus-infected person in a simpler manner.

[0100] When features are used as factors, the prediction model construction unit 213 constructs a prediction model by treating some or all of lymphocytes, ferritin, hemoglobin, C-reactive protein, and lactate dehydrogenase, among the multiple components contained in blood, as factors that are more important than other components contained in blood. This makes it possible to accurately predict the severity of illness in virus-infected individuals by treating test values ​​for components that are highly correlated with severity as important factors.

[0101] When some of the test values ​​for each of the multiple components extracted as features for an infected individual are missing, the feature acquisition means interpolates the missing values ​​using a representative value based on the test values ​​of other infected individuals. This allows, for example, for an infected person to be predicted, if a test value for any of the blood components is missing, this can be interpolated, and together with the test values ​​for the other components that are not missing, a prediction can be made regarding the severity of the virus infection.

[0102] [Realization of functions through hardware and software] The function of executing the series of processes according to the above-mentioned embodiment can be realized by hardware, software, or a combination of these. In other words, it is sufficient that the function of executing the series of processes described above is realized in any of the aggravation prediction systems S, and there is no particular limitation on how this function is realized.

[0103] For example, when the function of executing the above-mentioned series of processes is realized by a processor that executes arithmetic processing, the processor that executes this arithmetic processing includes not only processors that are composed of various types of processing devices alone, such as single processors, multiprocessors, and multicore processors, but also processors that combine these various types of processing devices with processing circuits such as ASICs (Application Specific Integrated Circuits) or FPGAs (Field-Programmable Gate Arrays).

[0104] Furthermore, for example, when the function of executing the above-mentioned series of processes is realized by software, the program constituting the software is installed in a computer via a network or a recording medium. In this case, the computer may be a computer in which dedicated hardware is built in, or may be a general-purpose computer (e.g., general electronic devices such as general-purpose personal computers) capable of executing a predetermined function by installing a program. Furthermore, the step of writing the program may include only processes that are performed chronologically according to the order, but may also include processes that are performed in parallel or individually. Furthermore, the step of writing the program may be performed in any order within the scope of the gist of the present invention.

[0105] A recording medium on which such a program is recorded may be provided to a user by being distributed separately from a computer main body, or may be provided to a user in a state in which the program is preinstalled in the computer main body. In this case, the recording medium distributed separately from the computer main body is composed of a magnetic disk (including a floppy disk), an optical disk, a magneto-optical disk, or the like. The optical disk is composed of, for example, a CD-ROM (Compact Disc-Read Only Memory), a DVD (Digital Versatile Disc), or a Blu-ray (registered trademark) Disc. The magneto-optical disk is composed of, for example, an MD (Mini Disc). These recording media are realized as, for example, a removable medium 51 or a removable medium 52, and are mounted in the drive 18 or the drive 28 and installed in the computer main body. In addition, the recording medium provided to a user in a state in which the program is preinstalled in the computer main body is composed of, for example, a ROM 12 or a ROM 22 on which a program is recorded, or an SSD (Solid State Drive) or a hard disk included in the storage unit 15 or the storage unit 25. [Explanation of symbols]

[0106] 10 Test data management device, 20 Aggravation prediction device, 11, 21 CPU, 12, 22 ROM, 13, 23 RAM, 14, 24 Communication unit, 15, 25 Memory unit, 16, 26 Input unit, 17, 27 Output unit, 18, 28 Drive, 51, 52 Removable media, 111 Test data management unit, 112 Test data provision unit, 151, 251 Test data memory unit, 211 Test data acquisition unit, 212 Feature extraction unit, 213 Prediction model construction unit, 214 Prediction execution unit, 252 Feature memory unit, 253 Prediction model memory unit, S Aggravation prediction system

Claims

1. A feature extraction means for extracting test values ​​of each of a plurality of components contained in the blood of a person infected with a virus as features; a model construction means for constructing a prediction model for predicting the onset of severity by performing a logistic regression analysis using the feature amount extracted from the blood of the subject infected person as a factor and using whether the subject infected person has become severe due to infection with the virus as a response variable; A prediction means for predicting the severity of the infected person who is a prediction target based on the feature amount extracted from the blood of the infected person who is a prediction target and the prediction model; A severe condition prediction device comprising:

2. the model construction means constructs a prediction model by regarding some or all of lymphocytes, ferritin, hemoglobin, C-reactive protein, and lactate dehydrogenase, among the multiple components contained in the blood, as factors having a higher importance than other components contained in the blood, when the feature amounts are used as factors; The severe illness prediction device according to claim 1 .

3. when a part of the test values ​​of each of the plurality of components extracted as the feature values ​​for a certain infected person is missing, the feature extraction means interpolates the missing part using a representative value based on the test values ​​of other infected persons. The severe illness prediction device according to claim 1 or 2.

4. A feature extraction step of extracting test values ​​of each of a plurality of components contained in the blood of a person infected with a virus as features; a model construction step of constructing a prediction model for predicting the onset of severity by performing a logistic regression analysis using the feature amount extracted from the blood of the subject infected person as a factor and whether or not the subject infected person has become severely ill due to infection with the virus as a response variable; A prediction step of predicting the severity of the infected person who is a prediction target based on the feature amount extracted from the blood of the infected person who is a prediction target and the prediction model; A method for predicting progression, comprising:

5. A feature extraction function that extracts the test values ​​of multiple components contained in the blood of a virus-infected person as features; and a model construction function for constructing a prediction model for predicting the onset of severity by performing a logistic regression analysis using the feature amount extracted from the blood of the subject infected person as a factor and using whether the subject infected person has become seriously ill due to infection with the virus as a response variable; and A prediction function for predicting the severity of the infected person who is a prediction target based on the feature amount extracted from the blood of the infected person who is a prediction target and the prediction model; A program characterized by causing a computer to realize the above.

Citation Information

Patent Citations

  • Detection determination method, detection determination device, detection determination program, and device

    JP2019092505A