Medical document preparation support system, medical document preparation support method, and program

Abstract

To provide a medical document preparation support technique that can be intuitively used by a user including a doctor.SOLUTION: A medical document preparation support system 2 has: a keyword specification unit 22 that can specify a keyword included in a template 10 on the basis of the template 10 including a text that can be arbitrarily set by a user and the keyword written in a distinguishable manner from the text; an extraction unit 24 that can input a prompt instructing extracting information on the keyword from a medical record 30 of a patient to an extraction system 4, and acquire an output corresponding to the keyword from the extraction system 4; and a document preparation unit 25 that can prepare a medical document including the output corresponding to the keyword obtained from the extraction system 4 at a position corresponding to the keyword in the template 10.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to medical document creation support technology. [Background technology]

[0002] Doctors involved in patients' medical care are responsible for preparing medical documents such as medical information reports and discharge summaries. Medical information reports record the patient's medical history, symptoms, test results, and treatment progress so that doctors at the receiving medical institution can quickly and accurately understand the patient's medical history and current condition. Discharge summaries record the patient's medical history, progress during hospitalization, condition at the time of discharge, and future treatment plans to help them communicate with medical professionals other than the attending physician at the medical institution to which the attending physician belongs after discharge.

[0003] Medical documents such as medical information reports and discharge summaries are created by doctors while referring to the patient's medical records, including the patient's medical records and test results, etc. A system that supports the creation of medical documents is desired.

[0004] Patent Document 1 discloses a system that uses an artificial intelligence engine that is trained to input each patient's medical records and information for creating a referral letter, and output the suspected disease name, the location of the disorder, and whether further examination or surgery is necessary. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2018-13825 Summary of the Invention [Problem to be solved by the invention]

[0006] As in Patent Document 1, a system that performs specific learning on a model called an artificial intelligence engine so that specific information is output requires appropriate learning according to its operation.

[0007] When using templates that can be freely configured by the doctor who is the user, there are systems that require SQL statements to be included in the template to extract the necessary information from medical records, and systems that require settings for extracting the necessary information from medical records to be set separately from the template.However, there is a need for a system that can be used intuitively even by doctors who are not familiar with information and communication technology such as SQL and systems.

[0008] The present disclosure provides a medical document creation support technology that can be intuitively used by users such as doctors. [Means for solving the problem]

[0009] The medical document creation support system of the present disclosure includes a keyword identification unit capable of identifying keywords included in a template based on a template containing text that can be set arbitrarily by a user and keywords written in a manner that can be distinguished from the text; an extraction unit capable of inputting a prompt to an extraction system instructing the extraction system to extract information related to the keywords from a patient's medical record and obtaining output corresponding to the keywords from the extraction system; and a document generation unit capable of generating a medical document by including the output corresponding to the keywords obtained from the extraction system at a position in the template that corresponds to the keywords. [Effects of the Invention]

[0010] According to the present disclosure, it is possible to facilitate the creation of medical documents. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a block diagram showing a medical document creation support system according to an embodiment of the present invention; [Figure 2] FIG. 10 is a diagram illustrating an example of a template. [Figure 3] FIG. 1 illustrates an example of a medical record. [Figure 4]1A and 1B are diagrams illustrating an example of a prompt according to the present embodiment and an example of an output from an extraction system based on the prompt. [Figure 5] FIG. 10 is a diagram illustrating an example of a generated medical document. [Figure 6] FIG. 10 is a block diagram showing a medical document creation support system according to another embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0013] FIG. 1 is a block diagram showing a medical document creation support system 2 according to this embodiment. As shown in FIG. 1, the medical document creation support system 2 according to this embodiment includes a template acquisition unit 20, a medical record acquisition unit 21, a keyword identification unit 22, a prompt generation unit 23, an extraction unit 24, and a document generation unit 25. Each of these units (20 to 25) is realized by one or more processors of a computer reading a predetermined program. The medical document creation support system 2 uses an extraction system 4 to generate a medical document based on a template 10 and medical records 30, and can provide the medical document to users such as doctors.

[0014] The extraction system 4 is a machine learning model such as an LLM (Large Language Model) that can respond to prompts by generating output corresponding to the prompts from the patient's medical record 30. It is also called generative AI, which can generate new content (such as text). The extraction system 4 can be implemented, for example, by Microsoft Azure OpenAI, Google GCP Gemini / MedLM, or Anthropic Claude.

[0015] The template acquisition unit 20 acquires a template 10 from a template storage unit 11, such as a database that stores templates 10. The acquired template 10 may be either one specified by the user or a default template associated with the user. The template 10 is a model of a medical document, and is created in advance individually for each predetermined unit, such as for each document type or each medical department in a medical institution, and is stored in the template storage unit 11.

[0016] FIG. 2 is a diagram illustrating an example of a template 10. As shown in FIG. 2, the template 10 includes text that can be arbitrarily set by a user (e.g., a medical professional such as a doctor) and keywords that can be distinguished from the text. Examples of text include greetings, names of medical departments, names of doctors, and preferred expressions in medical documents. Keywords are words used to extract information from the medical record 30 and can be arbitrarily set by the user. For example, to extract "abdominal pain" from a medical record 30 containing the text "The patient complains of abdominal pain," arbitrary keywords such as "symptoms," "main complaint," and "symptoms" can be set. Even if arbitrary keywords are set, the extraction system 4 is likely to interpret the keywords and extract information related to the keywords. This enables intuitive use of the system. In this embodiment, to distinguish keywords from text, keywords are bracketed between a start tag "{" as an identifier indicating the start of the keyword and an end tag "}" as an identifier indicating the end of the keyword. However, this is not limited to this. An example of template 10 shown in Figure 2 includes seven keywords: "admission date," "surgery date," "operated eye," "disease name," "operative procedure," "progress," and "discharge date." All text in template 10 other than these seven keywords is not a keyword and is not used for prompts.

[0017] The medical record acquisition unit 21 shown in FIG. 1 acquires medical records 30 from a medical record storage unit 31, such as a database that stores medical records 30. The acquired medical records 30 may be either those specified by the user or default records associated with the user. The medical records 30 are information that includes text representing the patient's medical history, test results, and the like. For example, as illustrated in FIG. 3, the medical records 30 include text. FIG. 3 is a diagram showing an example of the medical records 30. The process of acquiring the template by the template acquiring unit 20 and the process of acquiring the medical record 30 by the medical record acquiring unit 21 may be performed in any order.

[0018] The keyword identification unit 22 shown in FIG. 1 identifies one or more keywords included in the template 10. In this embodiment, the part between tags (identifiers) is identified as a keyword. This can be identified by using regular expressions or a parser. From the template 10 shown in FIG. 2, seven keywords are identified: "admission date," "surgery date," "operated eye," "disease name," "operative procedure," "progress," and "discharge date."

[0019] The prompt generation unit 23 shown in FIG. 1 generates a prompt. The prompt is an instruction for instructing the extraction system 4 to extract information related to the keywords identified by the keyword identification unit 22 from the patient's medical record 30. The prompt is written in a format acceptable to the extraction system 4. The prompt includes an instruction section for the model. The prompt may further include input data for which the model is requested to respond and a format specification section for specifying an output format. FIG. 4 shows an example of a prompt according to this embodiment and an example of output from the extraction system 4 based on the prompt. As shown in the upper part of FIG. 4, the prompt includes an instruction section stating, "Please extract and display the following items from the following medical record: date of admission, date of surgery, operated eye, disease name, surgical procedure, progress, and date of discharge." The format specification section and input data are not shown in FIG. 4.

[0020] The extraction unit 24 shown in Fig. 1 is configured to input the prompts generated by the prompt generation unit 23 to the extraction system 4 and obtain output corresponding to the keywords from the extraction system 4. Specifically, the extraction unit 24 inputs prompts to the extraction system 4 and receives output from the extraction system 4. The lower part of Fig. 4 shows output from the extraction system 4 obtained from the medical record 30 illustrated in Fig. 3 using some of the prompts shown in the upper part of Fig. 4. As shown in the lower part of Fig. 4, some output is obtained for seven keywords.

[0021] The document generation unit 25 shown in FIG. 1 generates a medical document based on the template 10 and the output obtained from the extraction system 4. Specifically, the document generation unit 25 generates the medical document by replacing keywords in the template 10 with the output corresponding to the keywords obtained from the extraction system 4. FIG. 5 is a diagram showing an example of the generated medical document. As shown in FIG. 5, the keywords in the template 10 shown in FIG. 2 have been replaced with the output of the extraction system 4.

[0022] With this system, users, such as doctors, simply enter their preferred text into a template 10, which serves as a template for medical documents, and enter keywords (using identification tags in this embodiment) in areas where reference to the medical records 30 is necessary, in a distinguishable manner. The system then automatically generates a medical document by extracting information about the keywords from the medical records 30 and adding it to the template. The user can easily create medical documents by simply reviewing and correcting the automatically generated document, rather than creating the medical document from scratch. Compared to directly prompting the generation AI (extraction system 4) to create the entire medical document, such as "create a medical information report" or "create a discharge summary," the system pinpoints the extraction of only information related to the keywords desired by the user. This allows the user to omit information the user considers unnecessary (information other than information related to the keywords) and create medical documents in the format desired by the user. Furthermore, since users can intuitively set keywords for extracting information from medical records 30 in templates 10 along with text, this system can be easily used even without knowledge of how to use extraction systems 4 such as generative AI (prompt engineering) or other information processing technologies, making it possible to provide a highly convenient system. In addition, since an extraction system 4 such as a generative AI that can interpret text is used, there is an advantage that the medical records 30 do not need to be organized into a structured data set. In addition, it may be possible to reduce operational costs by utilizing an external, existing extraction system 4. Furthermore, if the medical document creation support system 2 can call the extraction system 4, it becomes possible to configure the medical document creation support system 2 without incorporating the extraction system 4, which allows for a loose relationship between the medical document creation support system 2 and the extraction system 4, and makes it possible to easily change the extraction system 4 used whenever a better extraction system 4 is released.

[0023] [Another embodiment] (A) In this embodiment, the extraction unit 24 is configured to be able to acquire output from one type of extraction system 4, but is not limited to this. For example, as shown in Fig. 6, the extraction unit 24 may be configured to be able to acquire output from multiple (multiple types of) extraction systems 4 based on a prompt. Fig. 6 is a block diagram showing a medical document creation support system 2 of another embodiment.

[0024] (B) In the case of (A) above, as shown in FIG. 6, the evaluation unit 26 may have an evaluation unit 26 capable of evaluating the output obtained from the extraction system 4, and the document generation unit 25 may decide whether to adopt the output obtained from the extraction system 4 based on the evaluation result by the evaluation unit 26, and may generate a medical document using the output that has been decided to be adopted. In a first example, the extraction unit 24 inputs a prompt containing multiple keywords to multiple extraction systems 4 (systems A, B, and C) and obtains outputs (A to C) from each system (A to C). The evaluation unit 26 evaluates each output (A to C) and calculates an evaluation value. The document generation unit 25 generates a medical document using the output with the highest (best) evaluation value. As an evaluation method, assuming that the prompt contains multiple keywords, the more keywords that are obtained as output, the higher (better) the evaluation value. For example, if the number of keywords with responses in output A from extraction system A is four, the number of keywords with responses in output B from extraction system B is five, and the number of keywords with responses in output C from extraction system C is three, it may be determined that output B from extraction system B has the highest evaluation value. As a second example, the extraction unit 24 inputs a prompt including multiple keywords to a first extraction system 4 (system A) and obtains output A from system A. The evaluation unit 26 evaluates whether output A satisfies a predetermined evaluation condition. If it is determined that output A satisfies the predetermined evaluation condition, the document generation unit 25 generates a medical document using output A. On the other hand, if it is determined that output A does not satisfy the predetermined evaluation condition, the extraction unit 24 inputs a prompt to a second extraction system 4 (system B) and obtains output B from system B. The evaluation unit 26 evaluates whether output B satisfies the predetermined evaluation condition. If it is determined that output B satisfies the predetermined evaluation condition, the document generation unit 25 generates a medical document using output B. On the other hand, if it is determined that output B does not satisfy the predetermined evaluation condition, the extraction unit 24 inputs a prompt to the next extraction system 4, and the above process may be repeatedly executed until the predetermined evaluation condition is satisfied or a predetermined termination condition is satisfied. According to the first or second example above, even if a certain extraction system 4 cannot properly extract information corresponding to a keyword, the possibility of generating an appropriate document can be increased by using multiple types of extraction systems 4.

[0025] (C) In the above embodiment, start and end tags are used as identifiers to distinguish keywords, but this is not limiting. For example, the expression [{admission date}] may be written as "<admission date>" or "#admission date." Any expression may be used as long as it allows the keyword to be distinguished from other text.

[0026] (D) In ​​the above embodiment, if the medical document creation support system 2 is capable of calling the extraction system 4, the medical document creation support system 2 may or may not have the extraction system 4.

[0027] (E) In the above embodiment, the document generation unit 25 replaces keywords in the template 10 with output corresponding to the keywords obtained from the extraction system 4, for example, replacing "disease name" with "glaucoma." However, this is not limited to this. For example, the keyword may be replaced and left as it is, such as replacing "disease name" with "disease name glaucoma," so that both the keyword and the output corresponding to the keyword are included in the medical document. As an example, the output may be written after the keyword, or the output may be written before the keyword. That is, including the mode of replacing and deleting the keyword and the mode of leaving the keyword, the document generation unit 25 may execute a process of including output corresponding to the keyword obtained from the extraction system 4 in a position corresponding to the keyword in the template 10 (before or after the keyword, for example).

[0028] (F) In the above embodiment, the document generating unit 25 replaces and deletes the identifiers that make the keywords distinguishable, but this is not limiting, and the identifiers may be left as they are.

[0029] [1] As described above, in the above embodiment, the medical document creation support system 2 may include a keyword identification unit 22 that can identify keywords included in the template 10 based on the template 10, which includes text that can be set arbitrarily by the user and keywords that are written in a way that can be distinguished from the text; an extraction unit 24 that can input a prompt to the extraction system 4 instructing it to extract information related to the keywords from the patient's medical record 30 and obtain output corresponding to the keywords from the extraction system 4; and a document generation unit 25 that can generate a medical document by including output corresponding to the keywords obtained from the extraction system 4 at a position in the template 10 that corresponds to the keyword. With this configuration, users such as doctors can simply write text of their choice into template 10, which serves as a model for medical documents, and clearly write keywords in the parts of medical records 30 where reference is required.This allows medical documents to be automatically generated with information about keywords supplemented from medical records 30, making it easy to create medical documents.

[0030] [2] As in the medical document creation support system 2 described in [1] above, the system may be provided with an evaluation unit 26 capable of evaluating the output obtained from the extraction system 4, the extraction unit 24 configured to be able to acquire output from each of the multiple extraction systems 4 based on a prompt, and the document generation unit 25 may decide whether or not to adopt the output obtained from the extraction system 4 based on the evaluation result by the evaluation unit 26, and generate a medical document using the output that has been decided to be adopted. According to this configuration, even if a certain extraction system 4 cannot properly extract information corresponding to a keyword, by using multiple types of extraction systems 4, the possibility of generating an appropriate document can be increased.

[0031] [3] As in the above embodiment, the medical document creation support method may include extracting keywords from a template 10 containing text that can be set arbitrarily by a user and keywords written in a manner that can be distinguished from the text, inputting a prompt to an extraction system 4 instructing the extraction system 4 to extract information related to the keywords from a patient's medical record 30, obtaining output corresponding to the keywords from the extraction system 4, and generating a medical document including the output corresponding to the keywords obtained from the extraction system 4 at a position in the template 10 that corresponds to the keywords.

[0032] [4] As in the above embodiment, the program may cause one or more programs to perform the following operations: extracting keywords from a template 10 containing text that can be arbitrarily set by a user and keywords written in a manner that can be distinguished from the text; inputting a prompt to the extraction system 4 instructing the extraction system 4 to extract information related to the keywords from the patient's medical record 30; obtaining output corresponding to the keywords from the extraction system 4; and generating a medical document by including the output corresponding to the keywords obtained from the extraction system 4 in a position in the template 10 corresponding to the keywords.

[0033] Although the embodiments of the present disclosure have been described above with reference to the drawings, the specific configurations should not be considered to be limited to these embodiments. The scope of the present disclosure is defined not only by the description of the above embodiments but also by the claims, and further includes all modifications within the meaning and scope of the claims.

[0034] The structures employed in the above-described embodiments can be employed in any other embodiment. The specific configurations of the components are not limited to the above-described embodiments, and various modifications are possible within the scope of the present disclosure. [Explanation of symbols]

[0035] 2: Medical document creation support system 4: Extraction system 10: Template 22: Keyword identification section 24:Extraction part 25: Document generation unit 26: Evaluation section 30: Medical records

Claims

1. a keyword specifying unit that can specify keywords included in a template based on a template including text that can be arbitrarily set by a user and keywords that are written in a manner that can be distinguished from the text; an extractor capable of inputting prompts to an extraction system to extract information related to the keywords from the patient's medical records and obtaining output from the extraction system corresponding to the keywords; a document generation unit capable of generating a medical document by including an output corresponding to the keyword obtained from the extraction system at a position in the template corresponding to the keyword; A medical document creation support system equipped with the above.

2. an evaluation unit capable of evaluating the output obtained from the extraction system; the extraction unit is configured to be able to acquire the outputs from the plurality of extraction systems based on the prompt; 2. The medical document creation support system according to claim 1, wherein the document generation unit determines whether to adopt the output obtained from the extraction system based on the evaluation result by the evaluation unit, and is capable of generating the medical document using the output that has been decided to be adopted.

3. extracting the keywords from a template including text that can be arbitrarily set by a user and keywords written in a manner that can be distinguished from the text; inputting prompts to an extraction system instructing it to extract information relating to the keywords from the patient's medical record and obtaining output from the extraction system corresponding to the keywords; generating a medical document by including an output corresponding to the keyword obtained from the extraction system at a position in the template corresponding to the keyword; A medical document creation support method, including:

4. extracting the keywords from a template including text that can be arbitrarily set by a user and keywords written in a manner that can be distinguished from the text; inputting prompts to an extraction system instructing it to extract information relating to the keywords from the patient's medical record and obtaining output from the extraction system corresponding to the keywords; generating a medical document by including an output corresponding to the keyword obtained from the extraction system at a position in the template corresponding to the keyword; A program that causes one or more processors to execute the above.

Images