Medical record creation support system and method for supporting creation of medical records

Abstract

To provide a medical records creation support apparatus and a method for supporting creation of medical records capable of outputting an official name of a term even when the term, which includes variations in expression, is input.SOLUTION: A server 1 according to one aspect of the present invention includes: an operation input unit 13 for receiving input of an input text; and a standardized text search unit 16 for predicting and outputting a standardized text corresponding to the input text on the basis of learned data obtained by learning the correspondence relationship between standardized texts that are official names of medical treatments and input texts that indicate the same treatments as those indicated by the standardized texts but are expressed differently from the standardized texts.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a medical record creation support system and a medical record creation support method. [Background technology]

[0002] Conventionally, dentists record the patient's oral condition, treatment details, etc. in a medical record after examining the patient. Dentists often create medical records using a medical record system, etc. For example, Patent Document 1 discloses a medical record input support program that identifies treatment details based on a combination of the name of a specified patient's disease by referring to the treatment history and identification information of the treatment device used in treating the patient, and inputs the identified treatment details into an electronic medical record as the current treatment details for the specified patient. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-33777 Summary of the Invention [Problem to be solved by the invention]

[0004] In dentistry, there are many different expressions (synonyms) that express the same meaning when describing procedures performed on patients. For example, "metal crown restoration (4 / 5 crown (premolar and molar))," "4 / 5 crown," and "4 / 5 Cro" all mean the same thing. Also, "vital crown preparation" and "vital PZ" mean the same thing. Furthermore, "combined impression," "combined imp," and "C-imp" all mean the same thing.

[0005] One of the reasons for the existence of such a large number of synonyms with different expressions is the widespread use of abbreviations for terms used on medical receipts. The official names of terms used on medical receipts are often long. For this reason, if the official names are written in medical records as they are, the readability of the medical records will decrease. Abbreviations of official names are also sometimes created to simplify communication between medical staff.

[0006] Which abbreviations or synonyms become established as "common sense" for an individual will vary depending on the educational institution to which the dentist belonged, such as dental school, and who the instructors were at the educational institution. Furthermore, even in the medical record systems used after becoming a dentist, the types of abbreviations and expressions used vary from manufacturer to manufacturer, and there is currently no uniformity in terminology.

[0007] These differences in expression can be thought of as "dialects" in medical terminology. Dental clinics are also places where people who use different "dialects" gather. Discrepancies in abbreviations or expressions between dentists can become an obstacle in recording medical records or in communication. While there have been many attempts to standardize these differences in terminology, no standardization has been achieved yet. Furthermore, these abbreviations and expression patterns tend to increase with each insurance reform.

[0008] The present invention has been made in consideration of the above situation, and an object of the present invention is to provide a medical record creation support device and a medical record creation support method that can output the official name of a term even when a term containing inconsistent expression is input. [Means for solving the problem]

[0009] A medical record creation support device according to one aspect of the present invention includes a terminal device to which an input text indicating the details of a procedure performed on a patient is input, and a medical record creation support device that supports the creation of a medical record based on the input text transmitted from the terminal device. The medical record creation support device according to one aspect of the present invention includes an input unit that accepts input of the input text, and a first expression text output unit that predicts and outputs the first expression text corresponding to the input text based on learning data that has learned the correspondence between a first expression text that is the official name of the procedure and a second expression text that is text indicating the same procedure as the procedure indicated by the first expression text but is expressed in a different way from the first expression text. [Effects of the Invention]

[0010] According to at least one aspect of the present invention, it is possible to provide a medical record creation support device and a medical record creation support method that can output the official name of a term even when the term contains an inconsistent expression. Problems, configurations, and effects other than those described above will become apparent from the following description of the embodiments. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a diagram showing an example of the schematic configuration of a medical record creation support system according to an embodiment of the present invention; [Figure 2] FIG. 2 is a block diagram showing an example of the configuration of each control system of the server and the terminal device according to an embodiment of the present invention. [Figure 3] FIG. 3 is a diagram showing an example of an erroneous recognition pattern contained in text after speech conversion according to the first embodiment of the present invention. [Figure 4] 1 is a table showing an example of correspondence between input texts and standardized texts in training data according to the first embodiment of the present invention. [Figure 5] FIG. 2 is a diagram showing an example of correspondence between a prompt according to the first embodiment of the present invention and standardized text output as a search result from a learning model. [Figure 6]FIG. 2 is a diagram showing an example of the configuration of a medical record creation support screen according to the first embodiment of the present invention. [Figure 7] FIG. 2 is a diagram showing an example of the configuration of a medical record creation support screen according to the first embodiment of the present invention. [Figure 8] FIG. 2 is a diagram showing an example of the configuration of a medical record creation support screen according to the first embodiment of the present invention. [Figure 9] 1 is a flowchart showing an example of the procedure of a medical record creation support process according to the first embodiment of the present invention. [Figure 10] 10 is a table showing an example of correspondence between input text and standardized text in training data used for training by a training model according to a second embodiment of the present invention. [Figure 11] FIG. 10 is a diagram showing an example of display of a medical record in a case where only text is displayed according to the second embodiment of the present invention. [Figure 12] FIG. 10 is a diagram showing an example of the configuration of a medical record creation support screen according to a second embodiment of the present invention. [Figure 13] FIG. 10 is a diagram showing an example of the configuration of a medical record creation support screen according to a second embodiment of the present invention. [Figure 14] FIG. 10 is a diagram showing an example of the configuration of a medical record creation support screen according to a second embodiment of the present invention. [Figure 15] 10 is a flowchart showing an example of the procedure of a medical record creation support process according to a second embodiment of the present invention. [Figure 16] FIG. 10 is a diagram showing an example of correspondence between input text and standardized text in training data according to a modified example of the first embodiment of the present invention. [Figure 17] FIG. 10 is a diagram showing an example of correspondence between a set of input texts and a set of standardized texts in training data according to a modified example of the second embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, examples of modes for carrying out the present invention (hereinafter referred to as "embodiments") will be described with reference to the accompanying drawings. Various numerical values ​​in the embodiments of the present invention are merely examples, and the present invention is not limited to the embodiments described below. Furthermore, in this specification and drawings, identical components or components having substantially the same functions will be given the same reference numerals, and redundant explanations will be omitted.

[0013] <Overview of the medical record creation support system> First, the configuration of a medical record creation support system 100 according to an embodiment of the present invention will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of the schematic configuration of the medical record creation support system 100.

[0014] As shown in Fig. 1, the medical record creation support system 100 includes a server 1 and terminal devices 2-1 to 2-n (n is a natural number equal to or greater than 2). The server 1 and each of the terminal devices 2-1 to 2-n are communicably connected via a network N. In the following description, when it is not necessary to distinguish between the terminal devices 2-1 to 2-n, they will be collectively referred to as "terminal device 2."

[0015] The server 1 (an example of a medical record creation support device) is installed in, for example, a cloud environment, and generates sentences to be entered into the medical record using text (hereinafter referred to as "input text") sent from multiple terminal devices 2, and sends the generated sentences to the terminal devices 2. More specifically, the server 1 first converts the input text (an example of second expression text) sent from multiple terminal devices 2, which contains variations in expression, into standardized text (an example of first expression text), which is its official name.

[0016] Text containing inaccurate expressions includes, as mentioned above, abbreviations of the official names of terms, synonyms, etc. Text containing inaccurate expressions also includes incorrect expressions, i.e., text containing inaccurate expressions due to misrecognition by the user (hereinafter referred to as "misrecognition patterns"). The server 1 transmits the converted standardized text to the terminal device 2 and presents it to the user, such as a dentist. The server 1 also generates text to be entered into the medical record using the standardized text selected by the user, and transmits the generated text to the terminal device 2.

[0017] The medical record described below is a dental treatment record, i.e., an "electronic medical record," that has been computerized using the medical record management function of the server 1. This electronic medical record is also an insurance medical record that records the treatment given to the patient and the insurance points. The contents of the medical record created by the server 1 according to this embodiment are sent to a receipt computer (not shown) and transcribed onto the medical receipt. The receipt computer may be provided within the server 1 according to this embodiment.

[0018] The terminal device 2 is a device installed in a dental clinic and is composed of a PC (Personal Computer), a mobile terminal such as a smartphone, etc. The terminal device 2 accepts text to be recorded in the medical record input by a user such as a dentist, and transmits the accepted text to the server 1. When the text to be recorded in the medical record is input by voice, the terminal device 2 converts the voice into text and transmits the converted text to the server 1.

[0019] Furthermore, when standardized text is transmitted from the server 1, the terminal device 2 displays the standardized text on the screen of the terminal device 2 and accepts instructions from the user. Instructions from the user include instructions to adopt or not adopt the standardized text. The terminal device 2 also transmits information on the correspondence between the user's instructions and the standardized text to the server 1. Furthermore, when text to be entered into the medical record is transmitted from the server 1, the terminal device 2 transcribes the text to be entered into the medical record into the medical record.

[0020] <Configuration of the control system for the medical record creation support system> Next, the configuration of each control system of the server 1 and the terminal device 2 that make up the medical record creation support system 100 will be described with reference to Fig. 2. Fig. 2 is a block diagram showing an example of the configuration of each control system of the server 1 and the terminal device 2.

[0021] [server] As shown in Figure 2, the server 1 includes a control unit 10 connected to a bus B1, a memory unit 11, a communication I / F (Interface) unit 12, an operation input unit 13, an output unit 14, a standardized text search unit 15, and a search result output unit 16. The control unit 10 is an arithmetic unit configured with a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, and a RAM (Random Access Memory) 103.

[0022] The CPU 101 reads out from the ROM 102 the program code of the software that realizes each function of the server 1 according to this embodiment, and loads it into the RAM 103 for execution. The server 1 may include a processing device such as an MPU (Micro-Processing Unit) instead of the CPU 101. Variables, parameters, etc. that are generated during the calculation process by the CPU 101 are temporarily written into the RAM 103.

[0023] The storage unit 11 may be, for example, a hard disk drive (HDD), a solid state drive (SSD), a flexible disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, or a non-volatile memory card. The storage unit 11 stores an operating system (OS), various parameters, and a program for causing the server 1 to function. The program is stored in the form of a computer-readable program code. The program for causing the server 1 to function may be stored in the ROM 102.

[0024] The CPU 101 sequentially executes operations in accordance with the program code. In other words, the ROM 102 or the storage unit 11 is an example of a computer-readable non-transitory recording medium that stores a program to be executed by a computer.

[0025] The communication I / F unit 12 (an example of an input unit) is configured by a communication device or the like that controls communication between the terminal device 2. The communication I / F unit 12 accepts input of input text sent from the terminal device 2. The communication I / F unit 12 also transmits the standardized text output from the search result output unit 16 to the terminal device 2.

[0026] The operation input unit 13 is configured by, for example, a mouse, a keyboard, and the like, and generates an operation signal in response to an operation by a user and supplies it to the CPU 101. The output unit 14 is, for example, a monitor configured with an LCD (Liquid Crystal Display) or the like. The operation input unit 13 and the output unit 14 may be integrated into a touch panel. The server 1 may also be configured without the operation input unit 13 and the output unit 14.

[0027] The standardized text search unit 15 (an example of a first expression text output unit) searches for a standardized text corresponding to the input text transmitted from the terminal device 2 and outputs the search result. The standardized text search unit 15 includes a prompt generation unit 151 and a learning model 152.

[0028] The prompt generation unit 151 uses the input text transmitted from the terminal device 2 to generate a prompt to be input to the learning model 152. An example of the generation of a prompt by the prompt generation unit 151 will be described in detail later with reference to FIG.

[0029] The learning model 152 is configured by a language model such as LLM (Large Language Models), etc. Note that the learning model 152 may be configured by supervised learning AI (Artificial Intelligence), etc. When the learning model 152 is configured using supervised learning AI, the classification algorithm used by the AI ​​can be, for example, an algorithm such as "Multi-Class Neural Networks" or "Random Forest."

[0030] If learning model 152 is configured using a supervised learning AI or the like instead of a large-scale language model, prompt generation unit 151 is not required. Alternatively, an unsupervised learning model may be used as learning model 152.

[0031] The learning data of the learning model 152 is correspondence information between abbreviations of formal terms, synonyms, or misrecognition patterns of terms that are expected to be included in the input text and the formal names of the terms (hereinafter also referred to as "standardized text"). When the text transmitted from the terminal device 2 is text converted into speech, the possibility that the text contains misrecognition patterns is particularly high. Therefore, in this embodiment, the learning data to be learned by the learning model 152 also includes information on misrecognition patterns.

[0032] By configuring the training data of the training model 152 in this way, even if a text containing an incorrect expression is input to the training model 152, the training model 152 can output, as a search result, a standardized text corresponding to the term that the text is assumed to represent. Examples of misrecognition patterns contained in the text after speech conversion will be described in detail with reference to FIG. 3.

[0033] The learning model 152 predicts (infers) a standardized text that is assumed to correspond to the input text based on the learning data. Then, the learning model 152 outputs the top 1 standardized text with the highest confidence in the inference or the top N standardized texts with the highest confidence (N is a natural number equal to or greater than 2) together with information on the confidence.

[0034] Even if there is no standardized text in the training data that exactly corresponds to the input text, the learning model 152 always outputs some search result together with confidence level information according to an algorithm based on information such as the distance between the texts. Therefore, unlike when a simple conversion table method is used, even if there is no matching standardized text, the search result will still be output. The confidence level is expressed, for example, as a value between 0 and 1.

[0035] Note that the learning of the learning model 152 using the learning data is additionally performed when, for example, a term is added due to an insurance revision or the like, or when the official name of a term is changed. By performing such learning, even when the insurance is revised, the learning model 152 can appropriately predict and output the standardized text corresponding to the input text.

[0036] Furthermore, the learning model 152 in this embodiment may be trained not only with training data in which abbreviations, synonyms, or misrecognition patterns of official terms are associated one-to-one with correspondence information on standardized text, but also with training data in which they are associated set-to-set (many-to-many). In the following, the form in which the learning model 152 learns learning data associated one-to-one will be described as a "first embodiment," and the form in which the learning model 152 learns learning data associated set-to-set will be described as a "second embodiment." Note that when there is no need to particularly distinguish between the first and second embodiments, they will be described as "embodiments." The search result output unit 16 outputs the information on the standardized text output from the learning model 152 to the terminal device 2 via the communication I / F unit 12 together with the information on the confidence level.

[0037] [Terminal Device] Next, the configuration of the control system of the terminal device 2 will be described with reference to FIG. The terminal device 2 includes a control unit 20, a memory unit 21, an operation input unit 22, an output unit 23, a communication I / F unit 25, a voice-to-text conversion unit 26, a display control unit 27, and a medical record transcription unit 28, all of which are connected to the bus B2.

[0038] The control unit 20 is a calculation device that includes a CPU 201, a ROM 202, and a RAM 203. The CPU 201 reads out program code of software that realizes each function according to the embodiment from the ROM 202, expands it in the RAM 203, and executes it. Alternatively, the CPU 201 directly reads out the program code from the ROM 202 and executes it as is. The terminal device 2 may include a processing device such as an MPU instead of the CPU 201. Variables, parameters, etc. that are generated during the calculation processing by the CPU 201 are temporarily written to the RAM 203.

[0039] The storage unit 21 may be, for example, an HDD, SSD, flexible disk, optical disk, magneto-optical disk, CD-ROM, CD-R, or non-volatile memory card. In addition to the OS and various parameters, the storage unit 21 also stores programs for causing the terminal device 2 to function, a medical record 211, and the like. The programs are stored in the form of computer-readable program codes. The programs for causing the terminal device 2 to function may also be stored in the ROM 202.

[0040] The CPU 201 sequentially executes operations in accordance with the program code. In other words, the ROM 202 or the storage unit 21 is used as an example of a computer-readable non-transitory recording medium that stores a program to be executed by a computer.

[0041] The operation input unit 22 is configured with, for example, a mouse, a keyboard, and the like, and generates an operation signal according to an input by the user and supplies it to the CPU 201. The output unit 23 is, for example, a monitor configured with an LCD or the like. For example, a medical record creation support screen Sc (see FIG. 8, etc.) that supports the creation of a medical record is displayed on the output unit 23. The operation input unit 22 and the output unit 23 may be configured integrally as a touch panel.

[0042] The voice input unit 24 is configured by, for example, a microphone, and receives voice input of text to be recorded in the medical record, spoken by the user. The communication I / F 25 is configured by a communication device that controls communication with the server 1, etc.

[0043] The speech-to-text conversion unit 26 is a functional block that provides speech recognition and speech-to-text conversion functions, commonly known as "S2T (Speech to Text)," and converts the speech input to the speech input unit 24 into text and outputs it. The display control unit 27 controls the display of the medical record creation support screen Sc and the like on the output unit 23. The medical record transcription unit 28 transcribes the formatted text sent from the server 1 into the medical record.

[0044] <First embodiment> [Examples of misrecognition patterns contained in text after speech conversion] Next, an example of a recognition error pattern contained in the text after speech conversion will be described below. Figure 3 shows an example of a recognition error pattern contained in the text after speech conversion.

[0045] Figure 3 (1) shows the patient's remarks and the dentist's comments in the medical record in text form. Figure 3 (2) shows an example of text obtained by converting the text described in (1) into a voice read aloud using a speech-to-text conversion function such as S2T, which is standard on the OS. Figure 3 (3) shows an example of text obtained by converting the text described in (1) into a voice read aloud using a speech-to-text conversion function such as medical-specific S2T. In Figures 3 (2) and (3), misrecognized parts are underlined.

[0046] In the example of conversion using a standard speech-to-text function shown in Figure 3(2), the rate of misrecognition patterns is relatively low in the part corresponding to the patient's statement. On the other hand, in the latter part of the text, which contains a lot of technical terms and is the dentist's opinion, "upper right six all tooth surfaces" is misrecognized as "upper right six all paper surfaces" and "C place" is misrecognized as "P place."

[0047] Even in the example of a medical-specific speech-to-text conversion function shown in Figure 3(3), there are several misrecognition patterns in the dentist's opinion section in the latter half, which contains many technical terms. For example, "upper right 6 all tooth surfaces" is misrecognized as "upper right 6 whole body," and "C treatment" is misrecognized as "invasive."

[0048] In this embodiment, such erroneous recognition patterns are also associated with the correct names of terms and are learned by the learning model 152. Therefore, according to this embodiment, even if the text converted by the speech-to-text conversion unit 26 of the terminal device 2 includes an erroneous recognition pattern, the learning model 152 outputs information on the correct name of the text.

[0049] [Example of correspondence between input text and standardized text in training data] Next, an example of correspondence between input text and standardized text in training data will be described with reference to Fig. 4. Fig. 4 is a diagram showing an example of correspondence between input text and standardized text in training data.

[0050] The left column of Fig. 4 shows input text, and the right column shows standardized text. As shown in Fig. 4, misrecognition patterns in the input text are underlined. Also, in Fig. 4, for example, the standardized text "Metal Crown Restoration (4 / 5 Crown (Premolar and Molar))" is associated with its abbreviations "4 / 5 Crown" and "4 / 5 Cro," as well as the misrecognition pattern "4 / 5 Crop."

[0051] In addition, the abbreviations "renimp" and "C-imp" and the misrecognition pattern "Rainy part" are associated with the "associated impression" in the standardized text. Furthermore, the abbreviation "P-sho" and the misrecognition pattern "T-shop" are associated with the "periodontal disease treatment" in the standardized text.

[0052] The misrecognition patterns associated with the standardized text include not only terms that are actually misrecognized by the speech-to-text conversion unit 26, but also terms that are incorrect expressions and terms that are assumed to be incorrect expressions. By including these terms in the misrecognition patterns, even when a user such as a dentist inputs a term that is an expression that is incorrectly recognized, the learning model 152 outputs the correct name of the term as the standardized text.

[0053] [Example of generating prompts] Next, an example of prompt generation by prompt generation unit 151 will be described with reference to Fig. 5. Fig. 5 is a diagram showing an example of correspondence between a prompt and standardized text output as a search result from learning model 152.

[0054] The left column of FIG. 5 shows example prompts, and the right column shows example standardized text as a result of the learning model 152 search. The left column of FIG. 5 shows an example of a prompt generated by the prompt generation unit 151 when, for example, the term input from the terminal device 2 is "four-fifths of a volume." The prompt reads, "You are an experienced dentist in Japan. Based on the attached learning data, please suggest a correction to the text input by the user. Note that the text to be presented must be the "standardized text" from the learning data. If conversion is not required, "No conversion" should be output. Text input by the user: four-fifths of a volume."

[0055] When such a prompt is input, the learning model 152 configured with LLM can output "metal crown restoration (4 / 5 crown (premolar and molar))" as a search result based on the learning data shown in Figure 4.

[0056] [Example of medical record creation support screen configuration] Next, a configuration example of the medical record creation support screen Sc generated by the display control unit 27 of the terminal device 2 will be described with reference to Fig. 6 to Fig. 8. Fig. 6 to Fig. 8 are diagrams showing a configuration example of the medical record creation support screen Sc.

[0057] The medical record creation support screen Sc shown in Fig. 6 is a screen that is displayed when creating a medical record for patient X. The medical record creation support screen Sc shown in Fig. 6 displays a pop-up screen Scp that opens when an instruction to edit the medical record is given.

[0058] A menu display area Ar1 is displayed at the top of the pop-up screen Scp. The menu display area Ar1 is an area where buttons corresponding to various menus such as "Close," "Back," and "Next Item" are displayed.

[0059] Below the menu display area Ar1, a treatment input content display area Ar2 and a standardized text information display area Ar3 are displayed. The treatment input content display area Ar2 has the following fields: "item," "points," and "number of times." The "item" field shows the treatment item entered in the medical record, the "points" field shows the medical fee points, and the "number of times" field shows the number of times the treatment was performed. The treatment input content display area Ar2 in Figure 6 shows an example in which the "item" is "four-fifths of a volume," the points are "100," and the number of times is "1."

[0060] The standardized text information display area Ar3 displays the items of standardized text output as search results from the learning model 152 and their confidence levels. The standardized text display function is provided to the user when the user presses the standardized AI button Bn located at the right end of the menu display area Ar1.

[0061] Figure 7 shows an example of a screen that is displayed after detecting the press of the standardized AI button Bn. In the standardized text information display area Ar3 of the pop-up screen Scp shown in Figure 7, the item "Metal crown restoration (4 / 5 crowns (premolars and molars))" is displayed in the top row. The "Certainty" item indicates that the certainty is "0.96." In addition, the standardized text information display area Ar3 displays search results below Top 2, such as "Vital tooth crown formation" with a certainty of "0.001" and "ABO" with a certainty of "0.0002."

[0062] By checking the content displayed in the standardized text information display area Ar3, the user can appropriately select the standard text to be transcribed into the medical record while referring to the information on the degree of certainty.

[0063] When the user double-clicks (tap) on any of the items displayed in the standardized text information display area Ar3, the standardized text displayed in that item is selected as a term to be transcribed into the medical record. The selected standardized text is then transcribed into the medical record by the medical record transcription unit 28 (see FIG. 2). Note that the standardized text information display area Ar3 may be provided with additional buttons for selecting and confirming the standardized text, and transcription into the medical record may be performed when pressing of these buttons is detected.

[0064] Fig. 8 shows an example of the medical record creation support screen Sc after the standardized text has been transcribed. The medical record creation support screen Sc shown in Fig. 8 displays the standardized text selected by the user, i.e., "Metal crown restoration (4 / 5 crowns (premolar and molar))", transcribed into the medical record.

[0065] [Medical record creation support processing] Next, a medical record creation support process performed by the medical record creation support system 100 according to this embodiment will be described with reference to Fig. 9. Fig. 9 is a flowchart showing an example of the procedure of the medical record creation support process.

[0066] First, the speech input unit 24 (see FIG. 2) of the terminal device 2 of the medical record creation support system 100 accepts terms to be recorded in the medical record from the user by speech input (step SA1a) or by text input (step SA1b). If text is accepted by speech input in step SA1a, the speech-to-text conversion unit 26 converts the input speech into text (step SA2a).

[0067] After processing step SA2a or step SA1b, control unit 20 of terminal device 2 transmits the input text to server 1 via communication I / F unit 25 (step SA3). Next, communication I / F unit 12 of server 1 receives the input text transmitted from terminal device 2 (step SB1). Next, control unit 10 of server 1 inputs the input text received by communication I / F unit 12 to standardized text search unit 15 (step SB2).

[0068] Next, the prompt generation unit 151 of the standardized text search unit 15 generates a prompt using the input text and inputs the generated prompt to the learning model 152 (step SB3). The learning model 152 searches the learning data for a standardized text corresponding to the input text and outputs the search result (step SB4). Then, the search result output unit 16 transmits the standardized text and confidence level information output from the learning model 152 as the search result to the terminal device 2 via the communication I / F unit 12 (step SB5).

[0069] Next, the communication I / F unit 25 of the terminal device 2 receives the search results sent from the server 1 (step SA4). Then, the display control unit 27 of the terminal device 2 displays the received terminal results on the medical record creation support screen Sc (see FIG. 6) (step SA5). Next, the control unit 20 of the terminal device 2 determines whether any standardized text has been selected by the user on the medical record creation support screen Sc (step SA6). If it is determined in step SA6 that no standardized text has been selected (NO in step SA6), the control unit 20 repeats the determination in step SA6.

[0070] On the other hand, if it is determined in step SA6 that any standardized text has been selected (YES in step SA6), the medical record transcription unit 28 transcribes the standardized text selected by the user into the medical record (step SA7). After processing in step SA7, the medical record creation support process by the medical record creation support system 100 ends.

[0071] In the above-described embodiment, the standardized text search unit 15 of the server 1 predicts and outputs standardized text corresponding to the input text based on learning data that has learned the correspondence between standardized text, which is the official name of a procedure performed on a patient, and input text that includes abbreviations of the official name, synonyms, misrecognition patterns, etc. Therefore, according to this embodiment, even if a user such as a dentist inputs an abbreviation, synonym, or misrecognition pattern of a procedure into the terminal device 2, the official name of the procedure is output from the server 1.

[0072] Therefore, according to this embodiment, users such as dentists can comfortably write medical records using terms that they use on a daily basis. Furthermore, according to this embodiment, even if there is no standardization of the terms that indicate procedures, users can create medical records using terms that are unified (standardized).

[0073] Furthermore, in this embodiment, the standardized text search unit 15 predicts and outputs standardized text corresponding to the input text based on the training data. Therefore, a situation in which no standardized text corresponding to the input text exists, which is expected when the input text and the standardized text are associated with each other in a table or the like, cannot occur in this embodiment. In other words, according to this embodiment, even if the standardized text has a low degree of certainty in its correspondence with the input text, some standardized text can always be output from the server 1.

[0074] Furthermore, in this embodiment, the standardized text search unit 15 associates the standardized text with information on the degree of certainty of the correspondence with the input text and outputs the associated standardized text. The correspondence information between the standardized text and the degree of certainty is then displayed on the screen of the output unit 23 of the terminal device 2. Therefore, the user can select the standardized text to be transcribed into the medical record while referring to the degree of certainty information.

[0075] In addition, in this embodiment, the correspondence between input text, which is text converted from speech, and standardized text is also learned in the training data. This input text also includes text converted from speech due to erroneous recognition by the speech-to-text conversion unit 26. Therefore, even if the conversion by the speech-to-text conversion unit 26 has a specific quirk or the accuracy of the speech conversion is low, this embodiment makes it possible to create a medical record using the official names of terms.

[0076] <Second embodiment> In the first embodiment described above, an example was given in which the learning model 152 learns the correspondence between input text and standard text. In the second embodiment, an example will be described in which the learning model 152 learns a set of text. More specifically, the learning model 152 learns the correspondence between a set of term abbreviations, synonyms, and misrecognition patterns and a set of standardized text. In the second embodiment, the learning model 152 also learns a set of text that is missing input text corresponding to an action that should be included in a series of actions consisting of multiple actions, as a set of input text that includes misrecognition patterns.

[0077] [Example of correspondence between input text and standardized text in training data] Fig. 10 is a diagram showing an example of correspondence between input text and standardized text in the training data trained by the training model 152. As in Fig. 4, the left column of Fig. 10 shows input text, and the right column shows standardized text. Also, in the table shown in Fig. 10, misrecognition patterns are underlined.

[0078] In the top row of the table shown in Figure 10, a set of input texts, "4 / 5 crown," "BT (centric occlusion, paraffin wax used)," and "TeC," is associated with a set of standard texts, "Metal crown restoration (4 / 5 crown (premolar and molar))," "Combined impression," "BT (centric occlusion, paraffin wax used)," and "TeC."

[0079] "Metal crown restoration (4 / 5 crown (premolar and molar))" in the standardized text is the official name of the abbreviation "4 / 5 crown." On the other hand, "associated impression" in the standardized text is a term for which no corresponding abbreviation or synonym is included in the input text. The "associated impression" procedure should be included in a series of procedures that includes "metal crown restoration (4 / 5 crown (premolar and molar))," "BT (centric occlusion, using paraffin wax)," and "TeC" when these procedures are performed. Therefore, the reason why the "associative impression" procedure is not included in the series of procedures is thought to be that the user forgot to enter the item even though it was actually performed on the patient, or that the user did not understand the need to enter it (mistakenly believed that it was not necessary).

[0080] As described above, the learning model 152 in this embodiment learns the correspondence between a set of input texts generated taking into account omissions and misrecognitions, and a set of standard texts corresponding to multiple actions that should originally be performed as a series of actions. According to this embodiment, by performing such learning, even if a procedure that should be recorded in the medical record is missing from the set of text entered by the user, it is possible to fill in the procedure and present it to the user.

[0081] In the third row from the top of the table shown in Figure 10, a set of input texts "denture adjustment," "scaling," "cleaning," and "home hygiene instruction" is associated with a set of standard texts "denture adjustment," "scaling," "cleaning," "fluoride application," and "home hygiene instruction." In this example, the set of input texts consists of only official names. In the set of standardized texts corresponding to this set of input texts, the term "fluoride application" is added.

[0082] By matching the set of input texts with the set of standard texts in the learning data of learning model 152 in this way, it becomes possible to suggest to the user the treatment of "fluoride application" that should be performed along with the treatments of "denture adjustment," "scaling," "cleaning," and "visit hygiene instruction" as a treatment that should be included in the medical record.

[0083] [Example of generating prompts] Next, an example of prompt generation by prompt generation unit 151 will be described with reference to Fig. 11. Fig. 11 is a diagram showing an example of correspondence between a prompt and standardized text output as a search result from learning model 152.

[0084] 11, as in Fig. 5, the left column shows examples of prompts, and the right column shows examples of standardized text as search results of the learning model 152. That is, the left column of Fig. 11 shows examples of prompts generated by the prompt generation unit 151 when the set of text input from the terminal device 2 is "4 / 5 crown," "BT (centric occlusion, paraffin wax used)," and "TeC."

[0085] The prompt in Figure 11 reads, "You are an experienced dentist in Japan. Based on the attached training data, please suggest corrections to the text entered by the user. Please be sure to use the 'standardized text' from the training data when you present the text. Even if part of the set does not match, please find and present a set with similar text. Text set entered by the user: 4 / 5 crowns, BT (centric occlusion, paraffin wax used), TeC."

[0086] When such a prompt is input, the learning model 152 configured with LLM can output the search results "metal crown restoration (4 / 5 crowns (premolars and molars)), combined impression, BT (centric occlusion position, using paraffin wax) and TeC" based on the learning data shown in Figure 10.

[0087] [Example of medical record creation support screen configuration] Next, a configuration example of the medical record creation support screen Sc generated by the display control unit 27 of the terminal device 2 in the second embodiment will be described with reference to Fig. 12 to Fig. 14. Fig. 12 to Fig. 14 are diagrams showing a configuration example of the medical record creation support screen Sc.

[0088] The configuration of the medical record creation support screen Sc shown in FIG. 12 is the same as the configuration of the medical record creation support screen Sc shown in FIG. 6, so a description of the screen configuration will be omitted.

[0089] The standardized text information display area Ar3 of the pop-up screen Scp shown in Fig. 12 displays three items: "4 / 5 crown," "BT (centric occlusion, paraffin wax used)," and "TeC." When the user presses the standardized AI button Bn in this state, the set of input texts "4 / 5 crown," "BT (centric occlusion, paraffin wax used)," and "TeC" is sent to the server 1 and input into the standardized text search unit 15 of the server 1. Furthermore, under the control of the display control unit 27 of the terminal device 2, the screen transitions to the screen shown in Fig. 13.

[0090] 13, four items are displayed in the standardized text information display area Ar3 of the pop-up screen Scp: "Metal crown restoration (4 / 5 crowns (premolars and molars))," "Combined impression," "BT (centric occlusion, paraffin wax used)," and "TeC." These items are a set of standard texts output by the learning model 152 of the standardized text search unit 15 for the set of input texts described above.

[0091] The top row of the set of standard texts displayed in the standardized text information display area Ar3, "Metal crown restoration (4 / 5 crown (premolar and molar))," is the official name of "4 / 5 crown," one of the input texts. Also, the second row, "Associated impression," is the official name (standardized text) of a procedure that the learning model 152 has determined to be missing from the series of procedures that should have been included.

[0092] In addition, a "confidence" display area is provided to the right of the "item" display area in the standardized text information display area Ar3. The "confidence" display area displays information about the confidence of the set of standard texts displayed in the standardized text information display area Ar3. Figure 13 shows an example in which the confidence is "0.79."

[0093] In addition, below the display areas for "item" and "certainty" in the standardized text information display area Ar3, a comment display area Ar31 is provided, in which comments output from the learning model 152 are displayed. FIG. 13 shows an example in which the comment display area Ar31 displays the comment "AI comment: A missing item was found, so it was supplemented based on the learning data." By displaying such a display in the comment display area Ar31, the user can understand that there was a missing item in the item they entered.

[0094] Furthermore, below the comment display area, a content approval button Bn2 is provided. When the content approval button Bn2 is pressed, the multiple items displayed in the item display area of ​​the standardized text information display area Ar3 are transcribed to the medical record by the medical record transcription unit 28 of the terminal device 2.

[0095] Fig. 14 shows an example of the medical record creation support screen Sc after the standardized text has been transcribed. The medical record creation support screen Sc shown in Fig. 14 displays the standardized text selected by the user, i.e., "Metal crown restoration (4 / 5 crowns (premolar and molar))", "Combined impression", "BT (centric occlusion, paraffin wax used)", and "TeC", transcribed into the medical record.

[0096] [Medical record creation support processing] Next, the medical record creation support process by the medical record creation support system 100 according to this embodiment will be described with reference to Fig. 15. Fig. 15 is a flowchart showing an example of the procedure of the medical record creation support process.

[0097] The processes in steps SA11a, SA12a, and SA11b are the same as the processes in steps SA1a, SA2a, and SA1b in FIG. 9, and therefore will not be described again.

[0098] After processing step SA12a or step SA11b, the control unit 20 (see FIG. 2) of the terminal device 2 determines whether all text has been entered (step SA13). After entering all text, the user presses the standardized AI button Bn (see FIG. 12, etc.) on the medical record creation support screen Sc. Therefore, for example, if the control unit 20 detects that the standardized AI button Bn has been pressed, the determination in step SA13 is YES.

[0099] If it is determined in step SA13 that all the text has not been input (NO in step SA13), control unit 20 returns to step SA11a or step SA11b and waits for input from the user. On the other hand, if it is determined in step SA13 that all the text has been input (YES in step SA13), control unit 20 transmits the text to server 1 via communication I / F unit 25 (step SA14).

[0100] The processing from step SB11 to step SB15 performed in the server 1 after step SA14 is the same as the processing from step SB1 to step SB5 in Fig. 9. Furthermore, the processing from step SA15 to step SA18 performed in the terminal device 2 after step SB15 is the same as the processing from step SA4 to step SA7 in Fig. 9. Therefore, a description of these processing steps will be omitted.

[0101] The second embodiment described above provides the same advantages as the first embodiment. Furthermore, in the second embodiment, even if a user such as a dentist mistakenly recognizes a procedure that should be included in a series of procedures as unnecessary, or forgets to record a necessary procedure, the medical record creation support system 100 presents the results of the procedure that should have been recorded in the medical record being supplemented. Therefore, the user can become aware of their own mistake or omission and transcribe accurate information into the medical record.

[0102] Furthermore, according to the second embodiment, even if a set of input texts corresponding to procedures that are not supposed to be combined are mistakenly entered as a series of procedures, it is possible to prevent the incorrect combination of procedures from being transcribed directly into the medical record.

[0103] Furthermore, in the second embodiment, the learning model 152 searches for standardized texts using training data obtained by learning a set of input texts. Therefore, for example, even when text that the learning model 152 has not yet learned is input, the learning model 152 can search for a set of standardized texts that are assumed to correspond to the input texts, based on other input texts included in the set. The medical record creation support system 100 can then present the searched set of standardized texts to the user as suggested revisions to the term group to be transcribed into the medical record.

[0104] Currently, the definition of the set of procedures that should be recorded in the medical record varies from one medical record system to another. The accuracy of the definition of the set of procedures in a medical record system depends on the depth of insurance knowledge of the insurance technicians at the manufacturer of the medical record system. Furthermore, if a medical record system with an incorrect definition of the set of procedures that should be recorded in the medical record is used, there is a possibility that the procedures recorded in the medical record created by the medical record system will be incomplete or over-specified.

[0105] If there is an excess or deficiency in the treatment recorded on the receipt, the receipt will be returned by the insurer as a medical malpractice receipt. When a receipt is returned, the dental clinic must investigate and re-create the medical malpractice receipt, and the insurer will not pay the dental clinic medical expenses until the assessment is made after a re-examination. Therefore, if a medical record is created that shows an deficiency in a treatment that should have been included as a series of treatments, the dental clinic could suffer significant losses. Conversely, if a treatment that should not have been included as a series of treatments is recorded on the receipt, the patient will have to pay unnecessary medical expenses, which of course leads to an increase in the medical expenses paid by the government as medical fees.

[0106] According to the medical record creation support system 100 of the second embodiment, if there is a deficiency in the treatment recorded in the medical record, information about the deficiency is presented to the user. In other words, the medical record creation support system 100 of the second embodiment can also provide the user with a so-called "receipt check" function. Therefore, according to the second embodiment, it is possible to prevent the creation of a medical record that leads to an incomplete medical receipt.

[0107] In the first and second embodiments described above, the standard text or a set of standard texts output from the learning model 152 is first presented to the user as a revision proposal, and an instruction as to whether or not to transcribe the proposed text into the medical record is received from the user. However, the present invention is not limited to this. The medical record creation support system 100 according to this embodiment may automatically revise the content to be transcribed into the medical record based on the standard text or a set of standard texts output from the learning model 152.

[0108] In the above-described embodiment, an example was given in which the learning model 152 outputs standard texts that are the official names of abbreviations, synonyms, misrecognition patterns, etc. based on input texts that include these. In the second embodiment, an example was given in which the learning model 152 outputs a set of standard texts that are the official names of these, and a set of standard texts that correspond to the correct treatments, based on a set of input texts that include abbreviations, synonyms, misrecognition patterns, deficiencies, incorrect combinations, etc. However, the present invention is not limited to these examples. The medical record creation support system of the present invention may output standardized text in a second language different from the first language, based on input text in a first language.

[0109] Fig. 16 is a diagram showing an example of correspondence between input texts and standardized texts in training data according to a modification of the first embodiment, and Fig. 17 is a diagram showing an example of correspondence between a set of input texts and a set of standardized texts in training data according to a modification of the second embodiment. In each of the examples shown in Fig. 16 and Fig. 17, the first language is Japanese and the second language is English.

[0110] 16, for example, the English standardized text "partial veneer crown (4 / 5)" is associated with the Japanese abbreviations "4 / 5 kan" and "4 / 5 Cro" and the misrecognition pattern "five-fourth volume." By configuring the learning data in this way, even when Japanese (first language) terms including abbreviations, synonyms, misrecognition patterns, etc. are input, the medical record creation support system 100 can output standard texts that are the official names of these terms in English (second language).

[0111] 17, for example, a set of input texts of Japanese "4 / 5 crown," "BT (centric occlusion, paraffin wax used)," and "TeC" is associated with a set of standard texts of English "partial veneer crown (4 / 5)," "combined impression (C-imp)," "BT," and "temporary crown." By configuring the learning data in this way, the learning model 152 can output a set of standard texts that are the official names of these and a set of standard texts in English (second language) that correspond to the exact procedures, based on a set of input texts of Japanese (first language) that include abbreviations, synonyms, misrecognition patterns, deficiencies, incorrect combinations, etc.

[0112] Furthermore, in the above-described embodiment, the creation of medical records (electronic medical records) and medical receipts in dental treatment has been described, but the present invention can also be applied when creating medical records (electronic medical records) and medical receipts for treatment by doctors, nurses, physical therapists, judo therapists, etc. Furthermore, the medical record creation support system of the present invention can be applied not only to dentistry but also to various medical departments such as medicine, pharmacy, and veterinary medicine.

[0113] In the above-described embodiment, the object for which the medical record creation support system supports the creation of an insurance medical record required for claiming medical insurance claims was given as an example, but the present invention is not limited to this. The object for which the medical record creation support system of the present invention supports the creation of a "sub-medical record" containing various information obtained through communication with the patient may also be given as an example.

[0114] In addition, the above-mentioned embodiment examples provide detailed and specific explanations of the configurations of the devices (servers, terminal devices) and systems (medical record creation support systems) in order to clearly explain the present invention, and are not necessarily limited to those that have all of the configurations described.

[0115] 1 and 2, the control lines or information lines shown by solid lines are those considered necessary for explanation, and do not necessarily show all control lines or information lines in the product. In reality, it can be considered that almost all components are interconnected.

[0116] Furthermore, in this specification, processing steps describing chronological processing include not only processing that is performed chronologically in the order described, but also processing that is not necessarily performed chronologically but is performed in parallel or individually (for example, parallel processing or processing by objects). [Explanation of symbols]

[0117] 1...server, 2...terminal device, 2-1...terminal device, 2-n...terminal device, 15...standardized text search unit, 16...search result output unit, 25...speech to text conversion unit, 26...display control unit, 27...medical record transcription unit, 100...medical record creation support system, 151...prompt generation unit, 152...learning model, 153...search result output unit

Claims

1. A medical record creation support system including: a terminal device to which an input text indicating the content of a treatment performed on a patient is input; and a medical record creation support device that supports the creation of a medical record based on the input text transmitted from the terminal device, The medical record creation support device an input unit that accepts input of the input text; a first expression text output unit that predicts and outputs the first expression text corresponding to the input text based on learning data that has learned a correspondence between a first expression text that is a formal name of the action and a second expression text that is a text indicating the same action as the action indicated by the first expression text but is expressed in a different way from the first expression text. Medical record creation support system.

2. The second expression text includes at least one of an abbreviation of the official name, a synonym of the official name, and an inaccurate expression based on a misrecognition of the official name. The medical record creation support system according to claim 1.

3. The first-expression text output unit outputs information on a degree of confidence in a prediction that the second expression text corresponding to the input text corresponds to the first expression text, in association with the first expression text. The medical record creation support system according to claim 2.

4. When there are a plurality of first expression texts predicted to correspond to the input text, the first expression text output unit outputs correspondence information between the plurality of first expression texts and their respective confidence levels. The medical record creation support system according to claim 3.

5. The input text includes text converted from speech by a speech-to-text converter that converts speech into text. The medical record creation support system according to claim 4.

6. The training data is training data that has learned the correspondence between a set of the second expression texts of a plurality of different types and a set of the first expression texts. The medical record creation support system according to claim 4.

7. The inaccurate representation also includes a set of the second representation texts that does not include input text corresponding to an action that should be included in the set of actions that is made up of a plurality of the actions. The medical record creation support system according to claim 6.

8. the medical record creation support device further includes an output unit that transmits information of the first expression text output from the first expression text output unit to the terminal device; the terminal device includes a display control unit that displays information about the first expression text output from the output unit on a screen of the output unit; an operation input unit into which a user inputs instructions based on the content displayed on the screen; a chart transcription unit that transcribes the first expression text into the chart based on an instruction from the user input via the operation input unit. The medical record creation support system according to any one of claims 3 to 7.

9. A medical record creation support method using a medical record creation support system including a terminal device to which input text indicating the content of a treatment performed on a patient is input, and a medical record creation support device that supports the creation of a medical record based on the input text transmitted from the terminal device, an input unit of the medical record creation assistance device receiving input of the input text; a procedure in which a first expression text output unit predicts and outputs the first expression text corresponding to the input text based on learning data that has learned a correspondence between a first expression text that is a formal name of the action and a second expression text that is a text indicating the same action as the action indicated by the first expression text but is expressed in a different way from the first expression text. How to support medical record creation.

Images