Information processing device, information processing method and program
Patent Information
- Application Number
- JP2023139916
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-08-30
- Publication Date
- 2025-03-13
AI Technical Summary
【0011】 以上のように本開示によれば、医療診断で用いる決定木の生成を支援することができる。
Smart Images
Figure 2025033894000001_ABST
Abstract
Description
[Technical field]
[0001] The present disclosure relates to an information processing device and an information processing method for generating a decision tree used in medical diagnosis, and further relates to a program for implementing these. [Background technology]
[0002] In recent years, with the advancement of IT (Information Technology), a system has been proposed that presents the name of a patient's illness, treatment methods, the medical department that a patient should visit, and the like, simply by the patient answering questions on a terminal (for example, Non-Patent Document 1).
[0003] Specifically, the system disclosed in Non-Patent Document 1 presents multiple questions on a terminal regarding age, sex, past medical history, symptoms, etc. Then, when the system disclosed in Non-Patent Document 1 obtains sufficient answers to each question from the user, it presents possible disease names, treatment methods, the medical department to be visited, etc. based on the answers.
[0004] Moreover, the system disclosed in Non-Patent Document 1 is constructed with a chatbot using artificial intelligence (hereinafter also referred to as "AI" (Artificial Intelligence)). For this reason, the system disclosed in Non-Patent Document 1 inputs the answer entered by the user into a machine learning model to extract the meaning of the answer, and furthermore compares the extracted meaning with a decision tree created in advance, and determines the next question from the comparison result. This input of the answer and determination of the question are repeated until the final answer is determined. [Prior art documents] [Non-patent literature]
[0005] [Non-Patent Document 1] “Just answer a few questions to find out the name of the disease and where to go for treatment”, [online], Ubi, [Retrieved June 1, 2023], Internet<URL:https: / / ubie.app / > Summary of the Invention [Problem to be solved by the invention]
[0006] Incidentally, the creation of a decision tree is important when deriving an answer in an interactive format, such as in the system disclosed in the above-mentioned Non-Patent Document 1. However, since a decision tree is created manually by a specialist such as a doctor based on medical knowledge, there is a problem in that the creation of a decision tree is costly.
[0007] One example objective of the present disclosure is to aid in the generation of decision trees for use in medical diagnosis. [Means for solving the problem]
[0008] In order to achieve the above object, an information processing device according to one aspect of the present disclosure includes: a data extraction unit that inputs symptom data indicating a symptom into a learning model constructed by machine learning using symptom data indicating a symptom recognized by a patient and medical data indicating medical information corresponding to the symptom as training data, assigns weights to each element constituting the input symptom data, and extracts the element whose weight is equal to or greater than a threshold value; a candidate sentence generation unit that generates a branch condition candidate sentence using the extracted elements; a determination unit that determines, for each of the symptom data that is the training data, whether or not a symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating unit that generates a decision tree having the branching condition candidate sentences as nodes by using a determination result and the medical data that has become the training data; The present invention is characterized in that it is equipped with:
[0009] In order to achieve the above object, an information processing method according to one aspect of the present disclosure includes: a data extraction step of inputting symptom data indicating symptoms into a learning model constructed by machine learning using symptom data indicating symptoms recognized by a patient and medical data indicating medical information corresponding to the symptoms as training data, weighting each element constituting the input symptom data, and extracting the element whose weight is equal to or greater than a threshold value; a candidate sentence generating step of generating a branch condition candidate sentence using the extracted elements; a determining step of determining, for each of the symptom data that is the training data, whether or not the symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating step of generating a decision tree having the branching condition candidate sentences as nodes by using the determination result and the medical data that became the training data; The present invention is characterized in that it has the following features:
[0010] Furthermore, in order to achieve the above object, a program according to one aspect of the present disclosure includes: On the computer, a data extraction step of inputting symptom data indicating symptoms into a learning model constructed by machine learning using symptom data indicating symptoms recognized by a patient and medical data indicating medical information corresponding to the symptoms as training data, weighting each element constituting the input symptom data, and extracting the element whose weight is equal to or greater than a threshold value; a candidate sentence generating step of generating a branch condition candidate sentence using the extracted elements; a determining step of determining, for each of the symptom data that is the training data, whether or not the symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating step of generating a decision tree having the branching condition candidate sentences as nodes by using the determination result and the medical data that became the training data; The present invention is characterized in that: Effect of the Invention
[0011] As described above, according to the present disclosure, it is possible to assist in the generation of decision trees used in medical diagnosis. [Brief description of the drawings]
[0012] [Figure 1] FIG. 1 is a diagram illustrating a schematic configuration of an example of an information processing device. [Diagram 2] FIG. 2 is a diagram showing a more specific configuration of an example of the information processing device. [Diagram 3] FIG. 3 is a diagram illustrating an example of a learning model. [Figure 4] FIG. 4 is a diagram illustrating an example of training data. [Diagram 5] FIG. 5 shows an example of pre-processing performed on symptom data. [Figure 6] FIG. 6 is a diagram illustrating an example of the data extraction process and the process of generating a branch condition candidate sentence. [Figure 7] FIG. 7 is a diagram showing an example of a determination result as to whether or not a symptom includes a branch condition candidate statement. [Figure 8] FIG. 8 is a diagram showing an example of a decision tree generated from the determination result shown in FIG. [Figure 9] FIG. 9 shows an example of a screen displayed by the interactive judgment system. [Figure 10] FIG. 10 is a flow diagram illustrating an example of the operation of the information processing device. [Figure 11] FIG. 11 is a block diagram illustrating an example of a computer that realizes the information processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0013] (Embodiment) Hereinafter, an information processing device, an information processing method, and a program will be described in the embodiments with reference to FIGS.
[0014] [Device configuration] First, a schematic configuration of an example of an information processing device will be described with reference to Fig. 1. Fig. 1 is a configuration diagram showing a schematic configuration of an example of an information processing device.
[0015] An information processing device 10 shown in Fig. 1 is a device that generates a decision tree used in computer-based medical diagnosis. As shown in Fig. 1, the information processing device 10 includes a data extraction unit 11, a candidate sentence generation unit 12, a judgment unit 13, and a decision tree generation unit 14.
[0016] The data extraction unit 11 inputs symptom data indicating symptoms to the learning model, further weights each element constituting the input symptom data, and then extracts elements whose weights are equal to or greater than a threshold. The learning model is a model constructed by machine learning using symptom data indicating symptoms recognized by the patient and medical data indicating medical information corresponding to the symptoms as training data.
[0017] The candidate sentence generation unit 12 generates branching condition candidate sentences using the extracted elements. The judgment unit 13 judges, for each symptom data serving as training data, whether or not the symptoms indicated by the symptom data include each of the branching condition candidate sentences. The decision tree generation unit 14 uses the judgment result and the medical data serving as training data to generate a decision tree with the branching condition candidate sentences as nodes.
[0018] In this way, the information processing device 10 automatically generates a decision tree using a learning model for medical diagnosis and its training data without manual intervention. The information processing device 10 can assist in the generation of a decision tree used in medical diagnosis.
[0019] Next, the configuration and functions of the information processing device 10 will be specifically described with reference to Figures 2 to 9. Figure 2 is a configuration diagram showing the configuration of an example of an information processing device in more detail.
[0020] As shown in FIG. 2, the information processing device 10 includes a storage unit 15 and a learning model 20 in addition to the data extraction unit 11, the candidate sentence generation unit 12, the determination unit 13, and the decision tree generation unit 14 described above.
[0021] The learning model 20 is a model trained by machine learning so that when the subjective symptoms of a patient are input, the model outputs the medical department most suitable for diagnosing the patient, as shown in Fig. 3. Fig. 3 is a diagram showing an example of the learning model. The learning model 20 is a model built by machine learning using symptom data indicating the symptoms that the patient is aware of and medical data indicating medical information corresponding to the symptoms as training data.
[0022] Specifically, as shown in Fig. 4, the symptom data serving as training data is data that identifies the subjective symptoms described by the patient before medical treatment. Fig. 4 is a diagram showing an example of the training data. In the example of Fig. 4, the medical data is data indicating the name of the medical department that the patient who described the subjective symptoms actually visited.
[0023] As shown in Fig. 5, the symptom data is converted into a set of identifiers assigned to each word in a sentence indicating the symptom, and then input to the learning model 20. Specifically, text data indicating subjective symptoms constituting the symptom data is divided into words, and each word is converted into an identifier that is assigned to each word in advance. The obtained set of identifiers is used as the symptom data. Fig. 5 shows an example of pre-processing performed on the symptom data.
[0024] In the embodiment, the learning model 20 is constructed by a neural network having an attention mechanism, and weights are assigned by the attention mechanism. Specifically, since the symptom data is converted into a set of identifiers as described above and then input, the attention mechanism assigns weights to the identifiers.
[0025] Furthermore, the learning model 20 is actually implemented in the information processing device 10 by executing a machine learning program using the learned parameters. Note that the learning model 20 may be implemented in a device (computer) other than the information processing device 10.
[0026] As shown in Fig. 5 and Fig. 6, the data extraction unit 11 first divides the symptom data to be sample data into words, converts each word into an identifier, and inputs the resulting set of identifiers to the learning model 20. As a result, the learning model 20 assigns weights to each element (specifically, each identifier) constituting the symptom data by an attention mechanism. Then, as shown in Fig. 6, the data extraction unit 11 extracts elements from the learning model 20 whose weights are equal to or greater than a threshold value.
[0027] Specifically, in the example of FIG. 6, the data extraction unit 11 extracts the element "throat" which has the highest weight in the symptom data "uncomfortable feeling in the throat". FIG. 6 is a diagram showing an example of the data extraction process and the process of generating a branching condition candidate sentence. Note that elements are also referred to as "tokens" below. Also, the symptom data is actually a set of identifiers, and a weight is assigned to each identifier. However, in the example of FIG. 6, for ease of explanation, the symptom data and tokens are shown as text data.
[0028] As shown in Fig. 6, the candidate sentence generation unit 12 generates a branch condition candidate sentence by using the tokens (elements) extracted by the data extraction unit 11. In the example of Fig. 6, the data extraction unit 11 generates a branch condition candidate sentence by complementing the extracted token "nodo" with a sentence.
[0029] Specifically, the candidate sentence generation unit 12 first adds a blank to the extracted token. In FIG. 6, "'Nodo ga ■.'" is shown as the token to which a blank has been added. Next, the candidate sentence generation unit 12 selects an element to be put in ■ using, for example, an existing mask language model, and generates a branch condition candidate sentence. The number of branch condition candidate sentences generated may be one or more.
[0030] An example of an existing mask language model is BERT developed by Google. This language model calculates a score representing the possibility that each element constituting a vocabulary, which is prepared in advance, corresponds to a blank. The candidate sentence generation unit 12 selects an element whose score exceeds a threshold or the top N elements having the highest scores, and generates a branch condition candidate sentence. N is any natural number.
[0031] As shown in Fig. 7, for each symptom data serving as training data for the learning model 20, the determination unit 13 determines whether or not the symptoms indicating the symptom data include each branch condition candidate sentence. Fig. 7 is a diagram showing an example of a determination result as to whether or not a symptom includes a branch condition candidate sentence. The example in Fig. 7 also shows medical data corresponding to each symptom data. Also, in Fig. 7, if a symptom includes a branch condition candidate sentence, "includes" is displayed, and if not, "does not include" is displayed.
[0032] The decision tree generating unit 14 generates a decision tree having branching condition candidate sentences as nodes, using the judgment result by the judging unit 13 (see FIG. 7) and the medical data that is the training data, as shown in Fig. 8. Fig. 8 is a diagram showing an example of a decision tree generated from the judgment result shown in Fig. 7.
[0033] The decision tree generation unit 14 inputs the judgment result by the judgment unit 13 (see FIG. 7) and the medical data that became the training data into a function of a programming language having a function of generating a decision tree, and outputs a decision tree. As one method, the rpart function of the rpart library of the R language can be used. Specifically, the decision tree generation unit 14 extracts each column of the branching condition candidate sentences and the column of the medical department, which is the medical data, from the judgment result shown in FIG. 7, and converts "include" to "1" and "not include" to "0" in each column to create judgment data. Then, the decision tree generation unit 14 inputs the generated judgment data into the rpart function. As a result, the decision tree shown in FIG. 8 is output from the rpart function. The rpart function constructs a classification model or a regression model that can be expressed as a binary tree. The rpart function is also disclosed in the following reference document. References<URL:https: / / github.com / bethatkinson / rpart / >
[0034] The decision tree thus generated is used, for example, in an interactive judgment system that judges a medical department in an interactive manner. Fig. 9 shows an example of a screen displayed by the interactive judgment system. As shown in Fig. 9, a dialogue based on the decision tree is displayed, and when the patient inputs information in response to the dialogue, the medical department that the patient should visit is displayed.
[0035] [Device operation] Next, the operation of the information processing device 10 will be described with reference to Fig. 10. Fig. 10 is a flow diagram showing an example of the operation of the information processing device. In the following description, Figs. 1 to 9 will be referred to as appropriate. In addition, in the first embodiment, an information processing method is implemented by operating the information processing device 10. Therefore, in the embodiment, the description of the information processing method will be replaced with the following description of the operation of the information processing device 10.
[0036] As shown in FIG. 10, first, the data extraction unit 11 inputs symptom data as sample data into the learning model 20, and further assigns weights to each token constituting the input symptom data, and extracts elements whose weights are equal to or greater than a threshold value (step A1).
[0037] Next, the candidate sentence generating unit 12 generates branch condition candidate sentences using the tokens (elements) extracted by the data extracting unit 11 (step A2).
[0038] Next, the determination unit 13 determines, for each symptom data that is training data for the learning model 20, whether or not the symptom indicating the symptom data includes each branch condition candidate sentence (step A3).
[0039] Thereafter, the decision tree generating unit 14 uses the determination result in step A3 (see FIG. 7) and the medical data that has become the training data to generate a decision tree having the branching condition candidate sentences as nodes (step A4).
[0040] The decision tree generated by steps A1 to A4 is used in an interactive determination system that determines a medical department in an interactive manner, as described above. In this case, as shown in Fig. 9, a dialogue based on the decision tree is displayed, and when the patient inputs information in response to the dialogue, the medical department that the patient should visit is displayed.
[0041] In this way, the information processing device 10 automatically generates a decision tree using a learning model for medical diagnosis and its training data without manual intervention. The information processing device 10 can assist in the generation of a decision tree used in medical diagnosis.
[0042] [program] In the embodiment, the program may be a program that causes a computer to execute steps A1 to A4 shown in Fig. 10. By installing and executing this program in a computer, the information processing device 10 and the information processing method can be realized. In this case, the processor of the computer functions as a data extraction unit 11, a candidate sentence generation unit 12, a determination unit 13, and a decision tree generation unit 14 to perform processing.
[0043] The storage unit 15 may 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, or may be realized by a storage device of another computer. Examples of the computer include a general-purpose PC, a smartphone, and a tablet terminal device.
[0044] In the embodiment, the program 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 data extraction unit 11, the candidate sentence generation unit 12, the determination unit 13, and the decision tree generation unit 14.
[0045] [Physical configuration] Here, a computer that realizes the information processing device 10 by executing the program in the embodiment will be described with reference to Fig. 11. Fig. 11 is a block diagram showing an example of a computer that realizes the information processing device.
[0046] 11, a computer 110 includes a CPU (Central Processing Unit) 111, 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.
[0047] Furthermore, the computer 110 may include a GPU (Graphics Processing Unit) or an FPGA (Field-Programmable Gate Array) in addition to or instead of the CPU 111. In this embodiment, the GPU or FPGA can execute the programs in the embodiments.
[0048] 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).
[0049] 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.
[0050] 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.
[0051] 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.
[0052] Specific examples of the recording medium 120 include general-purpose semiconductor storage devices such as CF (Compact Flash (registered trademark)) and SD (Secure Digital), magnetic recording media such as a flexible disk, or optical recording media such as a CD-ROM (Compact Disk Read Only Memory).
[0053] The information processing device 10 can be realized not by a computer with a program installed, but by hardware corresponding to each part, for example, an electronic circuit. Furthermore, the information processing device 10 may be partially realized by a program and the remaining part by hardware. In the embodiment, the computer is not limited to the computer shown in FIG.
[0054] A part or all of the above-described embodiment can be expressed by (Additional Notes 1) to (Additional Notes 12) described below, but is not limited to the following descriptions.
[0055] (Appendix 1) a data extraction unit that inputs symptom data indicating a symptom into a learning model constructed by machine learning using symptom data indicating a symptom recognized by a patient and medical data indicating medical information corresponding to the symptom as training data, assigns weights to each element constituting the input symptom data, and extracts the element whose weight is equal to or greater than a threshold value; a candidate sentence generation unit that generates a branch condition candidate sentence using the extracted elements; a determination unit that determines, for each of the symptom data that is the training data, whether or not a symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating unit that generates a decision tree having the branching condition candidate sentences as nodes by using a determination result and the medical data that has become the training data; An information processing device comprising:
[0056] (Appendix 2) The learning model is constructed of a neural network having an attention mechanism, and the weights are assigned by the attention mechanism. 2. An information processing device according to claim 1.
[0057] (Appendix 3) The symptom data indicating symptoms that the patient is aware of is composed of a set of identifiers assigned to each word in a sentence indicating the symptoms, The attention mechanism assigns the weight to the identifier. 3. An information processing device according to claim 2.
[0058] (Appendix 4) The medical information corresponding to the symptom is information indicating a medical department corresponding to the symptom. 2. An information processing device according to claim 1.
[0059] (Appendix 5) a data extraction step of inputting symptom data indicating symptoms into a learning model constructed by machine learning using symptom data indicating symptoms recognized by a patient and medical data indicating medical information corresponding to the symptoms as training data, weighting each element constituting the input symptom data, and extracting the element whose weight is equal to or greater than a threshold value; a candidate sentence generating step of generating a branch condition candidate sentence using the extracted elements; a determining step of determining, for each of the symptom data that is the training data, whether or not the symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating step of generating a decision tree having the branching condition candidate sentences as nodes by using the determination result and the medical data that became the training data; 13. An information processing method comprising:
[0060] (Appendix 6) The learning model is constructed of a neural network having an attention mechanism, and the weights are assigned by the attention mechanism. 6. An information processing method as described in Appendix 5.
[0061] (Appendix 7) The symptom data indicating symptoms that the patient is aware of is composed of a set of identifiers assigned to each word in a sentence indicating the symptoms, The attention mechanism assigns the weight to the identifier. 6. An information processing method as described in Appendix 6.
[0062] (Appendix 8) The medical information corresponding to the symptom is information indicating a medical department corresponding to the symptom. 6. An information processing method as described in Appendix 5.
[0063] (Appendix 9) On the computer, a data extraction step of inputting symptom data indicating a symptom into a learning model constructed by machine learning using symptom data indicating a symptom recognized by a patient and medical data indicating medical information corresponding to the symptom as training data, weighting each element constituting the input symptom data, and extracting the element whose weight is equal to or greater than a threshold value; a candidate sentence generating step of generating a branch condition candidate sentence using the extracted elements; a determining step of determining, for each of the symptom data that is the training data, whether or not the symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating step of generating a decision tree having the branching condition candidate sentences as nodes by using the determination result and the medical data that became the training data; A program to execute.
[0064] (Appendix 10) The learning model is constructed of a neural network having an attention mechanism, and the weights are assigned by the attention mechanism. 9. The program described in Appendix 9.
[0065] (Appendix 11) The symptom data indicating symptoms that the patient is aware of is composed of a set of identifiers assigned to each word in a sentence indicating the symptoms, The attention mechanism assigns the weight to the identifier. 11. The program according to claim 10.
[0066] (Appendix 12) The medical information corresponding to the symptom is information indicating a medical department corresponding to the symptom. 9. The program described in Appendix 9. [Industrial Applicability]
[0067] The present disclosure is useful for interactive computer systems such as chatbots. [Explanation of symbols]
[0068] 10. Information processing device 11 Data Extraction Section 12 Candidate sentence generation section 13 Judgment section 14 Decision tree generator 15 Storage section 20 Learning Model 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 data extraction unit that inputs symptom data indicating a symptom into a learning model constructed by machine learning using symptom data indicating a symptom recognized by a patient and medical data indicating medical information corresponding to the symptom as training data, assigns weights to each element constituting the input symptom data, and extracts the element whose weight is equal to or greater than a threshold value; a candidate sentence generation unit that generates a branch condition candidate sentence using the extracted elements; a determination unit that determines, for each of the symptom data that is the training data, whether or not a symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating unit that generates a decision tree having the branching condition candidate sentences as nodes by using a determination result and the medical data that has become the training data; An information processing device comprising:
2. The learning model is constructed of a neural network having an attention mechanism, and the weights are assigned by the attention mechanism. The information processing device according to claim 1 .
3. The symptom data indicating symptoms that the patient is aware of is composed of a set of identifiers assigned to each word in a sentence indicating the symptoms, The attention mechanism assigns the weight to the identifier. The information processing device according to claim 2 .
4. The medical information corresponding to the symptom is information indicating a medical department corresponding to the symptom. The information processing device according to claim 1 .
5. a data extraction step of inputting symptom data indicating symptoms into a learning model constructed by machine learning using symptom data indicating symptoms recognized by a patient and medical data indicating medical information corresponding to the symptoms as training data, weighting each element constituting the input symptom data, and extracting the element whose weight is equal to or greater than a threshold value; a candidate sentence generating step of generating a branch condition candidate sentence using the extracted elements; a determining step of determining, for each of the symptom data that is the training data, whether or not the symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating step of generating a decision tree having the branching condition candidate sentences as nodes by using the determination result and the medical data that became the training data; 13. An information processing method comprising:
6. The learning model is constructed of a neural network having an attention mechanism, and the weights are assigned by the attention mechanism. The information processing method according to claim 5.
7. The symptom data indicating symptoms that the patient is aware of is composed of a set of identifiers assigned to each word in a sentence indicating the symptoms, The attention mechanism assigns the weight to the identifier. The information processing method according to claim 6.
8. The medical information corresponding to the symptom is information indicating a medical department corresponding to the symptom. The information processing method according to claim 5.
9. On the computer, a data extraction step of inputting symptom data indicating symptoms into a learning model constructed by machine learning using symptom data indicating symptoms recognized by a patient and medical data indicating medical information corresponding to the symptoms as training data, weighting each element constituting the input symptom data, and extracting the element whose weight is equal to or greater than a threshold value; a candidate sentence generating step of generating a branch condition candidate sentence using the extracted elements; a determining step of determining, for each of the symptom data that is the training data, whether or not the symptom indicated by the symptom data includes each of the branch condition candidate sentences; a decision tree generating step of generating a decision tree having the branching condition candidate sentences as nodes by using the determination result and the medical data that became the training data; A program to execute.