DIAGNOSIS SUPPORT DEVICE, DIAGNOSIS SUPPORT METHOD, AND PROGRAM

Abstract

A diagnosis assisting device (10) is provided with: a learning model selecting unit (11) for selecting, on the basis of information about a subject being diagnosed, one of first learning models that are generated in accordance with an individual attribute and indicate the relationship between an electrocardiogram waveform and a disease; an estimating unit (12) for analyzing, using the selected first learning model and on the basis of electrocardiogram data of the subject being diagnosed, the likelihood of the disease in the subject being diagnosed; and a presenting unit (13) for presenting an estimation result and the grounds for drawing the estimation result.

Description

[Technical field]

[0001] The present invention relates to a diagnostic support device and a diagnostic support method for supporting doctors in diagnosing heart diseases, and further relates to a program for implementing these. To Regarding. [Background technology]

[0002] An electrocardiogram is a graphical record of the electrical activity of a patient's heart. Doctors read the waveforms recorded on an electrocardiogram and use them to diagnose heart disease in patients.

[0003] However, it is not easy to find abnormalities from an electrocardiogram and depends on the skill of the doctor, so there is a possibility that the diagnosis results will vary. For this reason, Patent Document 1 discloses an analysis device that analyzes an electrocardiogram and outputs the analysis results.

[0004] Specifically, the analysis device disclosed in Patent Document 1 acquires electrocardiogram data of a patient, and separates the acquired electrocardiogram data into waveform data for each heartbeat. Next, the analysis device disclosed in Patent Document 1 classifies each waveform data based on predetermined classification conditions to create a waveform group with similar characteristics. After that, the analysis device disclosed in Patent Document 1 performs statistical processing on the waveform group to obtain statistical values ​​such as the number of abnormal heartbeats, their percentage of the total, and the maximum and minimum heart rates for each waveform group, adds patient information to the obtained statistical values, and outputs them as analysis results. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2007-20799 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, the analysis device disclosed in Patent Document 1 does not present the possibility of heart disease in consideration of patient information. Therefore, even if a doctor uses the analysis results from the device to diagnose heart disease, there is a possibility that the diagnosis results will vary.

[0007] One example of the object of the present invention is to provide a diagnostic support device, a diagnostic support method, and a diagnostic method for solving the above problems, which can present diagnostic materials based on patient information in diagnosing a patient's heart disease. program The purpose of this invention is to provide [Means for solving the problem]

[0008] In order to achieve the above object, a diagnosis support device according to one aspect of the present invention comprises: A learning model selection unit that selects a first learning model that indicates a relationship between an electrocardiogram waveform and a disease in accordance with a patient to be diagnosed; An estimation unit that estimates a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed by using the selected first learning model; A presentation unit that presents a result of the estimation and a reason for the result of the estimation; Equipped with It is characterized by:

[0009] In order to achieve the above object, a diagnostic support method according to one aspect of the present invention comprises: A learning model selection step of selecting a first learning model indicating a relationship between an electrocardiogram waveform and a disease according to a patient to be diagnosed; an estimation step of estimating a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed using the selected first learning model; a presentation step of presenting a result of the estimation and a reason for the result of the estimation; having It is characterized by:

[0010] Further, in order to achieve the above object, in one aspect of the present invention, program teeth, On the computer, A learning model selection step of selecting a first learning model indicating a relationship between an electrocardiogram waveform and a disease according to a patient to be diagnosed; an estimation step of estimating a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed using the selected first learning model; a presentation step of presenting a result of the estimation and a reason for the result of the estimation; Run R, It is characterized by: Effect of the Invention

[0011] As described above, according to the present invention, diagnostic materials based on patient information can be presented in diagnosing a patient's heart disease. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram showing a schematic configuration of a diagnosis support device according to an embodiment. [Diagram 2] FIG. 2 is a diagram explaining each wave in an electrocardiogram. [Diagram 3] FIG. 3 is a diagram specifically illustrating the configuration of a diagnosis support device according to an embodiment. [Figure 4] FIG. 4 shows an example of patient chart data used in the embodiment. [Diagram 5] FIG. 5 shows an example of electrocardiogram data of a patient used in the embodiment. [Figure 6] FIG. 6 is a diagram showing an example of electrocardiogram data and labels used as training data in the embodiment. [Figure 7] FIG. 7 is a diagram conceptually explaining the function of the second learning model (selection model) used in the embodiment. [Figure 8] FIG. 8 is a flow diagram showing the overall operation during machine learning processing in the diagnosis support device according to the embodiment. [Figure 9]FIG. 9 is a flow diagram specifically showing step A2 (learning process of the selection model) shown in FIG. [Figure 10] FIG. 10 is a diagram for explaining the process in step A21 shown in FIG. [Figure 11] FIG. 11 is a diagram for explaining the process in step A22 shown in FIG. [Figure 12] FIG. 12 is a flow diagram specifically showing step A3 (disease estimation model learning process) shown in FIG. [Figure 13] FIG. 13 is a diagram for explaining the process in step A33 shown in FIG. [Figure 14] FIG. 14 is a flow chart showing the operation of the diagnosis support process in the diagnosis support device in the embodiment. [Figure 15] FIG. 15 is a diagram for explaining the processing at each step shown in FIG. [Figure 16] FIG. 16 is a diagram showing an example of information presented to a user in the embodiment. [Figure 17] FIG. 17 is a block diagram illustrating an example of a computer that realizes a diagnosis support device according to an embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] (Embodiment) Hereinafter, a diagnosis support device, a diagnosis support method, and a program according to an embodiment will be described with reference to FIGS.

[0014] [Device configuration] First, a schematic configuration of a diagnosis support device in an embodiment will be described with reference to Fig. 1. Fig. 1 is a configuration diagram showing a schematic configuration of a diagnosis support device in an embodiment.

[0015] A diagnostic support device 10 according to an embodiment shown in Fig. 1 is a device for supporting a doctor in diagnosing a heart disease by using electrocardiogram data of a patient. As shown in Fig. 1, the diagnostic support device 10 includes a learning model selection unit 11, an estimation unit 12, and a presentation unit 13.

[0016] The learning model selection unit 11 selects a first learning model indicating a relationship between an electrocardiogram waveform and a disease according to a patient to be diagnosed. In the embodiment, the learning model selection unit 11 selects one of a plurality of learned models / estimated models as the first learning model based on information of the patient to be diagnosed. The first learning model is, for example, a machine-learned model regarding the relationship between electrocardiogram data and a disease, which is generated according to the attributes of the patient. Hereinafter, the first learning model is referred to as a "disease estimation model." The patient information includes personal information of the patient, for example, medical record data, biological data, and attribute information.

[0017] The estimation unit 12 estimates the possibility of a disease in a patient to be diagnosed based on the electrocardiogram data of the patient, using the disease estimation model selected by the learning model selection unit 11. The presentation unit 13 presents the result of the estimation by the estimation unit 12 and the basis for deriving the result of the estimation.

[0018] In this way, the diagnostic support device 10 selects a learning model suitable for a patient to be diagnosed from information of the patient to be diagnosed, such as age, sex, medical history, family history, and smoking history, and applies the electrocardiogram data of the patient to the selected learning model to estimate the possibility of disease. In other words, the diagnostic support device 10 can present diagnostic materials based on the information of the patient to be diagnosed in diagnosing a patient's heart disease.

[0019] Here, a general diagnosis of heart disease by a doctor using electrocardiogram data will be described with reference to FIG. 2. FIG. 2 is a diagram for explaining each wave in an electrocardiogram. As shown in FIG. 2, an electrocardiogram usually includes characteristic waveforms such as P waves, Q waves, R waves, S waves, T waves, and ST segments. A doctor reads the P waves, Q waves, R waves, S waves, T waves, and ST segments from the electrocardiogram, finds abnormalities in the state of each wave, and diagnoses heart disease. For example, if the T waves are flatter than normal or point downward from the baseline, the doctor will diagnose the patient as possibly having ischemic heart disease (angina pectoris, myocardial infarction) while checking the information in the patient's medical record.

[0020] In contrast, by using the diagnosis support device 10, the possibility of disease and its reason are presented from the electrocardiogram data in consideration of the patient's information (medical record) using an analysis model selected according to the patient to be diagnosed. Therefore, the possibility of variation in the diagnosis results by doctors is reduced.

[0021] Next, the configuration and functions of the diagnosis support device 10 in the embodiment will be specifically described with reference to Fig. 3 to Fig. 6. Fig. 3 is a configuration diagram specifically showing the configuration of the diagnosis support device in the embodiment.

[0022] 3, in the embodiment, the diagnosis support device 10 includes a learning model generation unit 14 and a storage unit 15 in addition to the above-mentioned learning model selection unit 11, estimation unit 12, and presentation unit 13. Also, as shown in FIG. 3, a display device 20 is connected to the diagnosis support device 10.

[0023] Furthermore, although not shown in Fig. 3, the diagnosis support device 10 is connected to an external device via a network so as to be able to communicate data. The external device transmits training data 30 used by the learning model generation unit 14, information (medical record data, etc.) 40 of the patient to be diagnosed, and electrocardiogram data 50 of the patient to be diagnosed, to the diagnosis support device 10. Fig. 4 shows an example of the medical record data of the patient used in the embodiment. Fig. 5 shows an example of the electrocardiogram data of the patient used in the embodiment.

[0024] The learning model generation unit 14 performs machine learning using the training data 30 to generate the disease estimation model 17. The machine learning method is not particularly limited. An example of the machine learning method is deep learning.

[0025] The training data includes previously acquired patient information, electrocardiogram data of the patient, and a label indicating the disease corresponding to the electrocardiogram data (hereinafter referred to as a "correct label"). Note that the "patient" in the training data is the patient from whom the training data was acquired. Patient information that serves as training data also includes the medical record data, biological information, and attribute information shown in FIG. 4. Patient electrocardiogram data that serves as training data also includes the electrocardiogram data shown in FIG. 5.

[0026] As shown in FIG. 6, the correct label indicating the disease corresponding to the electrocardiogram data may be a label given to each section of the electrocardiogram data. FIG. 6 is a diagram showing an example of electrocardiogram data and labels used as training data in the embodiment. In the example of FIG. 6, the sections are set by dividing the electrocardiogram data at a predetermined time interval. Hereinafter, each section will be referred to as "partial electrocardiogram data". In addition, in the example of FIG. 6, "normal", "atrial fibrillation", and "noise" are set as correct labels for each section of the electrocardiogram data. Examples of labels are not limited to the above examples, and other examples include "bigeminy", "arrhythmia", "myocardial infarction", and "angina pectoris". Furthermore, the labels may be classified into more detailed categories. For example, angina pectoris is classified into exertional angina pectoris, rest angina pectoris, vasospastic angina pectoris (atypical angina pectoris), arteriosclerotic angina pectoris, and asymptomatic myocardial ischemia.

[0027] In the embodiment, the learning model generation unit 14 first inputs the electrocardiogram data of the patient included in the training data 30 to the disease estimation model 17 and acquires an output result. Then, the learning model generation unit 14 performs machine learning using the acquired output result, the patient information, and a correct answer label indicating the disease corresponding to the electrocardiogram data as training data to generate a second learning model 16 indicating the correspondence between the patient information (medical record data) and the disease estimation model.

[0028] The second learning model 16 is used to select a disease estimation model 17 by the learning model selection unit 11, as described later. Hereinafter, the second learning model is referred to as a "selected model." The machine learning method in this case is not particularly limited. An example of the machine learning method is deep learning.

[0029] Fig. 7 is a diagram conceptually explaining the function of the second learning model (selection model) used in the embodiment. As shown in Fig. 7, when, for example, medical record data is input as patient information to the selection model 16, a disease inference model 17 is selected from disease inference models (1) to (M) according to the contents of the medical record data. M indicates the number of disease inference models prepared.

[0030] Furthermore, the learning model generation unit 14 can update the disease estimation model 17 using the selection model 16. Specifically, the learning model generation unit 14 first inputs information of individuals serving as training data into the selection model 16 to identify a disease estimation model corresponding to each individual. Then, the learning model generation unit 14 selects electrocardiogram data and a correct answer label corresponding to the identified disease estimation model from among electrocardiogram data of the patient serving as training data and correct answer labels indicating the disease corresponding to the electrocardiogram data. After that, the learning model generation unit 14 updates the disease estimation model using the selected electrocardiogram data and correct answer label.

[0031] Specifically, when the learning model generation unit 14 identifies a disease estimation model for each patient from which training data is obtained, it assigns electrocardiogram data and a correct label as training data for each patient. Then, the learning model generation unit 14 inputs the assigned electrocardiogram data for each patient to the identified disease estimation model, compares the output result with the correct label, and updates the disease estimation model based on the comparison result.

[0032] In the embodiment, the learning model selection unit 11 uses the above-mentioned selection model to select a disease estimation model suitable for the patient to be diagnosed from among disease estimation models generated in advance, based on information of the patient to be diagnosed.

[0033] In the embodiment, the estimation unit 12 analyzes the possibility of a disease of the patient to be diagnosed based on the output result of the disease estimation model for each partial electrocardiogram data obtained by dividing the electrocardiogram data of the patient to be diagnosed at a predetermined time interval. Specifically, the estimation unit 12 inputs each partial electrocardiogram data to the disease estimation model and obtains the output result. Then, the estimation unit 12 estimates the possibility of a disease of the patient to be diagnosed using the output result of each partial electrocardiogram data.

[0034] In the embodiment, the presentation unit 13 presents the result of the estimation and the grounds for deriving the result of the estimation on the screen of the display device 20. The result of the estimation presented at this time includes information on a disease that may occur in the patient who is the subject of diagnosis. The grounds presented at this time include the reason why the disease that may occur in the patient who is the subject of diagnosis was identified.

[0035] The basis may be at least one of partial electrocardiogram data (a specific portion) and patient attributes. Furthermore, the attributes include patient attributes corresponding to the selected disease estimation model. Furthermore, after presenting the basis, the presentation unit 13 can also present the result of the estimation in response to a request for presentation of the result of the estimation for the presented basis. Furthermore, the basis may be at least one of the data input to the disease estimation model. Examples of the basis include a waveform of a portion of the electrocardiogram waveform data, attributes of the patient to be diagnosed, etc.

[0036] [Device operation] Next, the operation of the diagnosis support device 10 in the embodiment will be described separately for machine learning processing and diagnosis support processing with reference to Figs. 8 to 16. In the following description, Figs. 1 to 7 will be referred to as appropriate. In the embodiment, a diagnosis support method is implemented by operating the diagnosis support device 10. Therefore, the description of the diagnosis support method in the embodiment will be replaced with the following description of the operation of the diagnosis support device 10.

[0037] Learning model generation process: First, the process of generating a learning model by the diagnosis support device 10 will be described with reference to Fig. 8 to Fig. 10. Fig. 8 is a flow diagram showing the overall operation during machine learning processing in the diagnosis support device in the embodiment.

[0038] In the embodiment, as a premise, in order to obtain training data 30, a patient to be diagnosed is interviewed in advance to create medical record data, and further, an electrocardiogram is measured on the patient to be diagnosed to obtain electrocardiogram data. In addition, a doctor diagnoses the electrocardiogram data, and a correct label is set for each section (see FIG. 6). That is, a correct label such as "normal", "atrial fibrillation", "bigeminy", or "noise" is assigned to the training data for each section.

[0039] As shown in FIG. 8, first, the learning model generation unit 14 sets the parameters of a model to be the selection model 16 and the parameters of a model to be the disease estimation model 17 to initial values ​​(step A1).

[0040] Next, the learning model generation unit 14 performs machine learning on the selection model indicating the correspondence between the medical record data and the disease estimation model (step A2). Specifically, the learning model generation unit 14 inputs the electrocardiogram data included in the training data 30 to the disease estimation model 17 and obtains an output result. Then, the learning model generation unit 14 performs machine learning using the obtained output result and the correct label included in the training data 30 as training data to update the parameters of the selection model 16.

[0041] Next, the learning model generation unit 14 executes machine learning to generate a disease estimation model 17 that indicates a relationship between an electrocardiogram waveform and a disease according to a patient (step A3). Specifically, the learning model generation unit 14 executes machine learning using the patient's medical record data, the patient's electrocardiogram data, and the correct answer label as training data, to update the parameters of the disease estimation model 17.

[0042] Next, the learning model generation unit 14 judges whether the number of times steps A2 and A3 have been executed has reached a predetermined number of times (step A4). If the result of the judgment in step A4 is that the number of times steps A2 and A3 have been executed has not reached the predetermined number of times, the learning model generation unit 14 executes step A2 again. On the other hand, if the number of times steps A2 and A3 have been executed has reached the predetermined number of times, the learning model generation unit 14 ends the machine learning process.

[0043] Next, the process in step A2 shown in Fig. 8 (learning process of selection model) will be specifically described with reference to Figs. 9 to 11. Fig. 9 is a flow diagram specifically showing step A2 (learning process of selection model) shown in Fig. 8. Fig. 10 is a diagram explaining the process in step A21 shown in Fig. 9. Fig. 11 is a diagram explaining the process in step A22 shown in Fig. 9.

[0044] As shown in FIG. 9, the learning model generation unit 14 inputs the electrocardiogram data of each patient from whom the training data 30 has been obtained to all the disease estimation models 17 (step A21).

[0045] For example, as shown in Fig. 10, the learning model generation unit 14 inputs the electrocardiogram data of each of the patients (1) to (N) from whom the training data was obtained to the disease estimation models (1) to (M). As a result, as shown in the bottom part of Fig. 10, the output results of the total disease estimation model 17 are obtained for each section of the electrocardiogram data.

[0046] In addition, in FIG. 10, the disease estimation model is written as "AI." K intervals are set in advance for the electrocardiogram data of each patient. N " is the interval ID (Identifier).

[0047] Next, the learning model generation unit 14 determines a disease estimation model 17 suitable for each patient from which the training data was obtained, based on the output result (step A22).

[0048] Specifically, in step A22, for each patient from whom training data was obtained, the learning model generation unit 14 identifies a disease inference model having a large number of intervals in which the estimation result is correct from among all disease inference models 17. Then, the learning model generation unit 14 determines the identified disease inference model as the disease inference model suitable for that patient.

[0049] Here, an example of the process of step A22 will be described. As shown in FIG. 11, the learning model generation unit 14 compares the correct label with the output result of step A21. Next, the learning model generation unit 14 creates an errata for each disease estimation model 17 for each section of the electrocardiogram data, and calculates the accuracy rate for each disease estimation model 17 for each patient using the created errata. Next, the learning model generation unit 14 identifies the disease estimation model 17 with the highest accuracy rate for each patient, and determines the identified disease estimation model 17 as the disease estimation model 17 suitable for that patient. Note that the process of step A22 is not limited to the above, and the disease estimation model may be determined based on a predetermined rule (patient attributes, biological information, medical record data, etc.).

[0050] Next, the learning model generation unit 14 regards the disease estimation model for each patient determined in step A22 as correct answer data, and performs machine learning using this correct answer data and the medical record data as training data to update the parameters of the selection model 16 (step A23). As a result, the selection model 16 is generated.

[0051] Next, the process in step A3 (disease estimation model learning process) shown in Fig. 8 will be specifically described with reference to Fig. 12 and Fig. 13. Fig. 12 is a flow diagram specifically showing step A3 (disease estimation model learning process) shown in Fig. 8. Fig. 13 is a diagram explaining the process in step A33 shown in Fig. 12.

[0052] As shown in FIG. 12, the learning model generation unit 14 first inputs the medical record data of each patient from whom training data has been obtained, into the selection model 16 (step A31).

[0053] Next, for each patient from whom training data has been obtained, the learning model generation unit 14 determines a disease estimation model 17 suitable for that patient based on the output result from the selection model 16 (step A32).

[0054] Next, for each patient from whom training data has been obtained, the learning model generation unit 14 assigns the electrocardiogram data of that patient as learning data for the determined disease estimation model (step A33).

[0055] 13, the learning model generation unit 14 identifies a corresponding patient for each disease estimation model 17, and allocates, for each disease estimation model 17, the electrocardiogram data of the patient corresponding to that disease estimation model 17. In the example of FIG. 13, the disease estimation model (1) is assigned electrocardiogram data 7-1 to 7-K7 of the corresponding patient (7) and electrocardiogram data 103-1 to 103-K of the corresponding patient (103). 103 and are allocated.

[0056] Next, the learning model generation unit 14 performs machine learning using the medical record data of the corresponding patient and the electrocardiogram data sorted in step A33 as training data for each disease estimation model, and updates the parameters of the disease estimation model 17 (step A34). As a result, the disease estimation model 17 is generated.

[0057] Diagnostic support process: The diagnosis support processing by the diagnosis support device 10 will be described with reference to Figs. 14 to 16. Fig. 14 is a flow diagram showing the operation of the diagnosis support processing by the diagnosis support device in the embodiment. Fig. 15 is a diagram explaining the processing at each step shown in Fig. 14. Fig. 16 is a diagram showing an example of information presented to a user in the embodiment.

[0058] 14, first, the learning model selection unit 11 acquires information (e.g., medical record data) 40 of a patient to be diagnosed (step B1). The learning model selection unit 11 also acquires electrocardiogram data of the patient to be diagnosed.

[0059] Next, the learning model selection unit 11 selects a disease estimation model 16 for estimating the possibility of a disease from the electrocardiogram data of the patient to be diagnosed, based on the patient information 40 acquired in step B1. Specifically, the learning model selection unit 11 selects a disease estimation model 17 suited to the patient from among disease estimation models 17 generated in advance, based on the output result of step B1 (step B2).

[0060] In addition, in step B2, the learning model selection unit 11 outputs the selected disease estimation model 17 and information on which the selection was based to the estimation unit 12. As shown in Fig. 15, the information on which the selection was based is the parameters and their values ​​used to select the disease estimation model 17 (see Fig. 7).

[0061] Next, the estimation unit 12 inputs the electrocardiogram data 50 of the patient to be diagnosed to the disease estimation model 17 selected in step B2, and estimates the possibility of the patient's disease (step B3). n The data is divided into sections and input (see FIG. 15).

[0062] For example, in step B3, the estimation unit 12 divides the electrocardiogram data 50 at a predetermined time interval to set sections. Then, as shown in Fig. 15, the estimation unit 12 inputs each piece of partial electrocardiogram data obtained by setting the sections to the disease estimation model 17 and calculates a certainty factor indicating the possibility of a disease for each section.

[0063] In the example of FIG. 15, in the section ID (1), the confidence level is calculated as 0.7 for atrial fibrillation and 0.2 for bigeminy. In addition, the output result by general deep learning may be used as the confidence level. The method of calculating the confidence level is not particularly limited.

[0064] Furthermore, the estimation unit 12 uses the certainty of each disease for each interval to calculate an overall certainty for each disease as a possibility of the patient's disease. Specifically, the estimation unit 12 may use the maximum certainty among the certainty for each interval as the overall certainty, or may use the average value of the certainty for each interval as the overall certainty. In addition, when the average value is calculated as the overall certainty, only the top several certainty values ​​with the highest values ​​may be used. The method of calculating the overall certainty is not particularly limited.

[0065] Next, the estimation unit 12 identifies the basis for the disease estimation in step B3 based on the information on the basis of the selection output from the learning model selection unit 11 (step B4). Specifically, the estimation unit 12 identifies, as the basis, the interval in which the certainty level was equal to or higher than a certain value for the cardiac disease in which the certainty level is equal to or higher than a threshold value.

[0066] Next, the estimation unit 12 outputs the possibility of the disease estimated in step B3 and the basis identified in step B4 to the presentation unit 13 (step B5). Specifically, the estimation unit 12 outputs to the presentation unit 13 the name of the heart disease for which the certainty is equal to or greater than a threshold, and the section ID of the section in which the certainty indicating the possibility of the heart disease was equal to or greater than a certain value.

[0067] Next, the presentation unit 13 presents the result and the basis of the estimation output in step B3 on the screen of the display device 20 (step B6). The presentation unit 13 can also present on the screen of the display device 20 the "information that is the basis of the selection" output to the estimation unit 12 in step B2.

[0068] For example, as shown in the example of Figure 16, the presentation unit 13 displays on the screen of the display device 20 the names of cardiac diseases for which the certainty factor is above a threshold as candidate diagnostic results, and displays partial electrocardiogram data for sections where the certainty factor was above a certain value as evidence points within the electrocardiogram.

[0069] 16, the presentation unit 13 also displays the "information on the basis of selection" output in step B3 as the basis location in the medical record on the screen of the display device 20. Furthermore, in the example of Fig. 16, when the user selects the name of one cardiac disease on the screen, the presentation unit 13 displays partial electrocardiogram data of the section corresponding to the selected cardiac disease.

[0070] The displayed basis is not limited to the above. For example, the basis may be the item name included in the medical record data of the target patient, the biological information of the target patient, or the like, or may be any other information that can be used to estimate a disease.

[0071] [Effects of the embodiment] As described above, in the embodiment, the patient's chart and electrocardiogram are used to perform estimation, and the possibility of heart disease is presented as the estimation result. In addition, in the embodiment, the basis for the estimation result and the relevant parts of the patient's chart are also presented. Therefore, according to the embodiment, in diagnosing a patient's heart disease using an electrocardiogram, diagnostic materials based on the patient's chart can be presented, and variation in the diagnosis result of heart disease is suppressed.

[0072] [program] An example of the program in the embodiment is a program that causes a computer to execute steps A1 to A4 shown in Fig. 8 and steps B1 to B6 shown in Fig. 14. By installing and executing this program in a computer, the diagnostic support device and diagnostic support method in the embodiment can be realized. In this case, the processor of the computer functions as a learning model selection unit 11, an estimation unit 12, a presentation unit 13, and a learning model generation unit 14, and performs processing.

[0073] The storage unit 15 can be realized by storing the data files constituting the storage unit 15 in a storage device such as a hard disk provided in the computer. Examples of the computer include a general-purpose PC, a smartphone, and a tablet terminal device.

[0074] The program in the embodiment may be executed by a computer system constructed by a plurality of computers. In this case, for example, each computer may function as any one of the learning model selection unit 11, the estimation unit 12, the presentation unit 13, and the learning model generation unit 14.

[0075] [Physical configuration] Here, a computer that realizes the diagnosis support device by executing the program in the embodiment will be described with reference to Fig. 17. Fig. 17 is a block diagram showing an example of a computer that realizes the diagnosis support device in the embodiment.

[0076] As shown in FIG. 17, a computer 110 includes a CPU (Central Processing Unit) 1 The computer includes a power supply 11, a main memory 112, a storage device 113, an input interface 114, a display controller 115, a data reader / writer 116, and a communication interface 117. These components are connected to each other via a bus 121 so as to be able to communicate data with each other.

[0077] The computer 110 may further include a GPU (Graphics Processing Unit) or an FPGA (Field-Programmable Gate Array) in addition to or instead of the CPU 111. In this aspect, the GPU or FPGA can execute the program in the embodiment.

[0078] The CPU 111 loads a program in the embodiment, which is composed of a group of codes and is stored in the storage device 113, into the main memory 112, and executes each code in a predetermined order to perform various calculations. The main memory 112 is typically a volatile storage device such as a DRAM (Dynamic Random Access Memory).

[0079] Moreover, the program in the embodiment is provided in a state stored in a computer-readable recording medium 120. The program in the embodiment may be distributed on the Internet connected via the communication interface 117.

[0080] Specific examples of the storage device 113 include a hard disk drive and a semiconductor storage device such as a flash memory. The input interface 114 mediates data transmission between the CPU 111 and input devices 118 such as a keyboard and a mouse. The display controller 115 is connected to a display device 119 and controls the display on the display device 119.

[0081] Data reader / writer 116 mediates data transmission between CPU 111 and recording medium 120, reads programs from recording medium 120, and writes processing results in computer 110 to recording medium 120. Communication interface 117 mediates data transmission between CPU 111 and other computers.

[0082] Specific examples of the recording medium 120 include a CF (Compact Flash (registered trademark)) and a Examples of such storage media include general-purpose semiconductor storage devices such as Secure Digital (D), magnetic recording media such as flexible disks, and optical recording media such as compact disk read only memories (CD-ROMs).

[0083] The diagnosis support device 10 in the embodiment can be realized by using hardware (e.g., circuits, etc.) corresponding to each part, instead of a computer with a program installed. Furthermore, the diagnosis support device 10 may be realized in part by a program and the remaining part by hardware.

[0084] A part or all of the above-described embodiment can be expressed by (Addendum 1) to (Addendum 36) described below, but is not limited to the following description.

[0085] (Appendix 1) A learning model selection unit that selects a first learning model that indicates a relationship between an electrocardiogram waveform and a disease in accordance with a patient to be diagnosed; An estimation unit that estimates a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed by using the selected first learning model; A presentation unit that presents a result of the estimation and a reason for the result of the estimation; Equipped with A diagnostic support device comprising:

[0086] (Appendix 2) 2. The diagnosis support device according to claim 1, the learning model selection unit selects one of the first learning models based on information of the patient to be diagnosed, using a second learning model indicating a correspondence relationship between information of the patient and the first learning models; A diagnostic support device comprising:

[0087] (Appendix 3) 3. The diagnosis support device according to claim 1, the estimation unit estimates a possibility of a disease of the patient to be diagnosed based on an output result of the first learning model for each partial electrocardiogram data obtained by dividing the electrocardiogram data of the patient to be diagnosed at a predetermined time interval; A diagnostic support device comprising:

[0088] (Appendix 4) 4. The diagnosis support device according to claim 3, The estimation unit analyzes a possibility of a disease of the patient to be diagnosed based on a result of analyzing each of the partial electrocardiogram data. A diagnostic support device comprising:

[0089] (Appendix 5) 5. The diagnosis support device according to claim 1, The result of the prediction includes the disease. A diagnostic support device comprising:

[0090] (Appendix 6) 6. The diagnosis support device according to claim 1, The rationale includes the reason why the disease was identified. A diagnostic support device comprising:

[0091] (Appendix 7) 7. The diagnosis support device according to claim 6, The basis is at least one of the electrocardiogram data or the attributes of the patient to be diagnosed. A diagnostic support device comprising:

[0092] (Appendix 8) 8. The diagnosis support device according to claim 7, the learning model selection unit selects the first learning model according to an attribute of the patient to be diagnosed; The attribute is an attribute corresponding to the selected first learning model. A diagnostic support device comprising:

[0093] (Appendix 9) 9. The diagnosis support device according to any one of claims 1 to 8, the presentation unit presents the grounds, and presents the result of the inference in response to a request for presentation of a result of the inference for the presented grounds. A diagnostic support device comprising:

[0094] (Appendix 10) 10. The diagnosis support device according to any one of claims 1 to 9, a learning model generation unit that generates the first learning model by machine learning using training data including information about an individual, electrocardiogram data of the individual, and a label indicating a disease corresponding to the electrocardiogram data; Equipped with A diagnostic support device comprising:

[0095] (Appendix 11) 11. The diagnosis support device according to claim 10, the learning model generation unit generates a second learning model by machine learning using an output result from the first learning model for the electrocardiogram data of the individual, information about the individual, and a label indicating a disease corresponding to the electrocardiogram data as training data; A diagnostic support device comprising:

[0096] (Appendix 12) 12. The diagnosis support device according to claim 11, The learning model generation unit is using the second learning model, identifying the first learning model corresponding to the individual based on information of the individual that constitutes the training data; updating the first learning model using the electrocardiogram data and the label selected as corresponding to the identified first learning model from among the electrocardiogram data of the individual that serves as the training data and the label indicating a disease corresponding to the electrocardiogram data; A diagnostic support device comprising:

[0097] (Appendix 13) A learning model selection step of selecting a first learning model indicating a relationship between an electrocardiogram waveform and a disease according to a patient to be diagnosed; an estimation step of estimating a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed using the selected first learning model; a presentation step of presenting a result of the estimation and a reason for the result of the estimation; having A diagnostic support method comprising:

[0098] (Appendix 14) 14. The diagnostic support method according to claim 13, before MemoirsIn a learning model selection step, a second learning model indicating a correspondence between patient information and the first learning models is used to select one of the first learning models based on information of the patient to be diagnosed. A diagnostic support method comprising:

[0099] (Appendix 15) The diagnostic support method according to claim 13 or 14, In the estimation step, a possibility of a disease of the patient to be diagnosed is estimated based on an output result of the first learning model for each partial electrocardiogram data obtained by dividing the electrocardiogram data of the patient to be diagnosed at a predetermined time interval. A diagnostic support method comprising:

[0100] (Appendix 16) 16. The diagnostic support method according to claim 15, In the estimation step, a possibility of a disease of the patient to be diagnosed is analyzed based on a result of analyzing each of the partial electrocardiogram data. A diagnostic support method comprising:

[0101] (Appendix 17) 17. The diagnostic support method according to any one of claims 13 to 16, The result of the prediction includes the disease. A diagnostic support method comprising:

[0102] (Appendix 18) 18. The diagnostic support method according to any one of claims 13 to 17, The rationale includes the reason why the disease was identified. A diagnostic support method comprising:

[0103] (Appendix 19) 19. The diagnostic support method according to claim 18, further comprising: The basis is at least one of the electrocardiogram data or the attributes of the patient to be diagnosed. A diagnostic support method comprising:

[0104] (Appendix 20) 20. The diagnostic support method according to claim 19, further comprising: In the learning model selection step, the first learning model is selected according to an attribute of the patient to be diagnosed, The attribute is an attribute corresponding to the selected first learning model. A diagnostic support method comprising:

[0105] (Appendix 21) 21. The diagnostic support method according to any one of appendix 13 to 20, In the presenting step, the grounds are presented, and in response to a request for presentation of a result of the inference on the presented grounds, a result of the inference is presented. A diagnostic support method comprising:

[0106] (Appendix 22) 22. The diagnostic support method according to any one of appendix 13 to 21, a learning model generation step of generating the first learning model by machine learning using training data including personal information, the personal electrocardiogram data, and a label indicating a disease corresponding to the electrocardiogram data; Equipped with A diagnostic support method comprising:

[0107] (Appendix 23) 23. The diagnostic support method according to claim 22, In the learning model generating step, a second learning model is generated by machine learning using an output result from the first learning model for the electrocardiogram data of the individual, information about the individual, and a label indicating a disease corresponding to the electrocardiogram data as training data. A diagnostic support method comprising:

[0108] (Appendix 24) 24. The diagnostic support method according to claim 23, In the learning model generation step, using the second learning model, identifying the first learning model corresponding to the individual based on information of the individual that constitutes the training data; updating the first learning model using the electrocardiogram data and the label selected as corresponding to the identified first learning model from among the electrocardiogram data of the individual that serves as the training data and the label indicating a disease corresponding to the electrocardiogram data; A diagnostic support method comprising:

[0109] (Appendix 25) On the computer, A learning model selection step of selecting a first learning model indicating a relationship between an electrocardiogram waveform and a disease according to a patient to be diagnosed; an estimation step of estimating a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed using the selected first learning model; a presentation step of presenting a result of the estimation and a reason for the result of the estimation; Run R, Program Hmm.

[0110] (Appendix 26) As described in Appendix 25 program And, before Memoirs In a learning model selection step, a second learning model indicating a correspondence between patient information and the first learning models is used to select one of the first learning models based on information of the patient to be diagnosed. Characterized by program .

[0111] (Appendix 27) The method described in Appendix 25 or 26 program And, In the estimation step, a possibility of a disease of the patient to be diagnosed is estimated based on an output result of the first learning model for each partial electrocardiogram data obtained by dividing the electrocardiogram data of the patient to be diagnosed at a predetermined time interval. Characterized by program .

[0112] (Appendix 28) As described in Appendix 27 program And, In the estimation step, a possibility of a disease of the patient to be diagnosed is analyzed based on a result of analyzing each of the partial electrocardiogram data. Characterized by program .

[0113] (Appendix 29) The method according to any one of claims 25 to 28. program And, The result of the prediction includes the disease. Characterized by program .

[0114] (Appendix 30) The method according to any one of claims 25 to 29. program And, The rationale includes the reason why the disease was identified. Characterized by program .

[0115] (Appendix 31) As described in Appendix 30 program And, The basis is at least one of the electrocardiogram data or the attributes of the patient to be diagnosed. Characterized by program .

[0116] (Appendix 32) As stated in Appendix 31 program And, In the learning model selection step, the first learning model is selected according to an attribute of the patient to be diagnosed, The attribute is an attribute corresponding to the selected first learning model. Characterized by program .

[0117] (Appendix 33) Any one of appendix 25 to 32 program And, In the presenting step, the grounds are presented, and in response to a request for presentation of a result of the inference on the presented grounds, a result of the inference is presented. Characterized by program .

[0118] (Appendix 34) Any one of appendix 25 to 33 program And, a learning model generation step of generating the first learning model by machine learning using training data including personal information, the personal electrocardiogram data, and a label indicating a disease corresponding to the electrocardiogram data; Equipped with Characterized by program .

[0119] (Appendix 35) As described in Appendix 34 program And, In the learning model generating step, a second learning model is generated by machine learning using an output result from the first learning model for the electrocardiogram data of the individual, information about the individual, and a label indicating a disease corresponding to the electrocardiogram data as training data. Characterized by program .

[0120] (Appendix 36) As stated in Appendix 35 program And, In the learning model generation step, using the second learning model, identifying the first learning model corresponding to the individual based on information of the individual that constitutes the training data; updating the first learning model using the electrocardiogram data and the label selected as corresponding to the identified first learning model from among the electrocardiogram data of the individual that serves as the training data and the label indicating a disease corresponding to the electrocardiogram data; Characterized by program .

[0121] Although the present invention has been described above with reference to the embodiment, the present invention is not limited to the above embodiment. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. [Industrial Applicability]

[0122] As described above, according to the present invention, in diagnosing a patient's heart disease, diagnostic materials based on the patient's information can be presented. The present invention is useful in a system for analyzing heart disease using an electrocardiogram. [Explanation of symbols]

[0123] 10. Diagnostic support device 11 Learning model selection section 12 Estimation part 13 Presentation section 14 Learning model generation unit 15 Storage section 16 Selection model (second learning model) 17 Disease Prediction Model (First Learning Model) 20 Display device 30 Training Data 40 Patient information (medical record data) 50 Patient ECG data 110 Computer 111 CPU 112 Main memory 113 Storage device 114 Input Interface 115 Display Controller 116 Data Reader / Writer 117 Communication Interface 118 Input Devices 119 Display device 120 Recording media 121 Bus

Claims

1. a learning model selection unit that uses a second learning model that indicates a correspondence between patient information and the first learning models from among a plurality of first learning models that indicate a relationship between electrocardiogram data and a disease, and selects one of the first learning models based on information of a patient to be diagnosed; An estimation unit that estimates a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed by using the selected first learning model; A presentation unit that presents a result of the estimation and a reason for the result of the estimation; Equipped with The first learning model is generated by machine learning using training data, which are information about a patient, electrocardiogram data of the patient, and a correct answer label indicating a disease corresponding to the electrocardiogram data; The second learning model is generated by machine learning using training data, which are an output result of the first learning model, patient information, and a correct answer label indicating a disease corresponding to the electrocardiogram data; The result of the prediction includes information about a disease that may be occurring in the patient to be diagnosed, The basis includes a reason why a disease that may be occurring in the patient to be diagnosed has been identified; A diagnostic support device comprising:

2. The diagnosis support device according to claim 1 , the estimation unit estimates a possibility of a disease of the patient to be diagnosed based on an output result of the first learning model for each partial electrocardiogram data obtained by dividing the electrocardiogram data of the patient to be diagnosed at a predetermined time interval; A diagnostic support device comprising:

3. The diagnosis support device according to claim 2, The estimation unit analyzes a possibility of a disease of the patient to be diagnosed based on a result of analyzing each of the partial electrocardiogram data. A diagnostic support device comprising:

4. The diagnosis support device according to claim 1 , The basis is at least one of the electrocardiogram data or the attributes of the patient to be diagnosed. A diagnostic support device comprising:

5. The diagnosis support device according to claim 4, The learning model selection unit selects the first learning model according to an attribute of the patient to be diagnosed, The attribute is an attribute corresponding to the selected first learning model. A diagnostic support device comprising:

6. 1. A computer-implemented method comprising: selecting one of a plurality of first learning models that indicate a relationship between electrocardiogram data and a disease, based on information of a patient to be diagnosed, using a second learning model that indicates a correspondence relationship between patient information and the first learning models; Using the selected first learning model, estimate a possibility of a disease of the patient to be diagnosed based on electrocardiogram data of the patient to be diagnosed; Present the results of the estimation and the grounds for the results of the estimation; The first learning model is generated by machine learning using training data, which are information about a patient, electrocardiogram data of the patient, and a correct answer label indicating a disease corresponding to the electrocardiogram data; The second learning model is generated by machine learning using training data, which are an output result of the first learning model, patient information, and a correct answer label indicating a disease corresponding to the electrocardiogram data; The result of the prediction includes information about a disease that may be occurring in the patient to be diagnosed, The basis includes a reason why a disease that may be occurring in the patient to be diagnosed has been identified; A diagnostic support method comprising:

7. On the computer, A step of selecting one of a plurality of first learning models that indicate a relationship between electrocardiogram data and a disease, based on information of a patient to be diagnosed, using a second learning model that indicates a correspondence relationship between patient information and the first learning models; Using the selected first learning model, a possibility of a disease of the patient to be diagnosed is estimated based on electrocardiogram data of the patient to be diagnosed; A step of presenting a result of the estimation and a reason for the result of the estimation, A program to be executed, The first learning model is generated by machine learning using training data, which are information about a patient, electrocardiogram data of the patient, and a correct answer label indicating a disease corresponding to the electrocardiogram data; The second learning model is generated by machine learning using training data, which are an output result of the first learning model, patient information, and a correct answer label indicating a disease corresponding to the electrocardiogram data; The result of the prediction includes information about a disease that may be occurring in the patient to be diagnosed, The grounds include a reason why a disease that may occur in the patient to be diagnosed was identified. program.

Images