Electronic case generation method, electronic device, and storage medium

Abstract

The application discloses an electronic case generation method, an electronic device and a storage medium. In the method, initial dialogue text corresponding to doctor-patient dialogue voice is acquired; for a plurality of case sub-topics, each case sub-topic is spliced with the initial dialogue text to determine corresponding updated dialogue text; each updated dialogue text is provided to a preset topic content generation model to determine answer content corresponding to each case sub-topic; and an electronic case is generated based on the answer content of each case sub-topic. Thus, the user does not need to locate the sub-title content in the doctor-patient dialogue, the content extraction and generation of different sub-titles can be realized based on a single model, the cost of generating the electronic case is reduced under the premise of ensuring the recognition effect.

Description

Technical Field

[0001] This invention belongs to the field of speech processing technology, and particularly relates to an electronic medical record generation method, an electronic device, and a storage medium. Background Technology

[0002] With the continuous development of internet technology, many hospitals now use electronic medical records to manage and maintain patient information. However, manual input by doctors leads to low efficiency and the loss of important information. Therefore, automatically generating electronic medical records based on speech recognition technology is a popular research direction in the industry.

[0003] In some current related technologies, when collecting patient case information, it is necessary to manually locate the source of each subheading (e.g., chief complaint, past medical history, diagnosis, etc.) matched by the voice input. Specifically, based on the patient's questions, the patient's chief complaint is summarized, or based on the doctor's questions and the patient's answers, the presence of certain past medical histories is summarized. Therefore, it is not only necessary to obtain the doctor-patient dialogue and the doctor's case report, but also to trace the content of each subheading in the case report back to its original source in the dialogue. This results in more sample annotation work, and it is a serial process, that is, the source needs to be located first, and then sufficient feature information from the corresponding source to generate the case content needs to be collected.

[0004] However, for a given subheading, there may be multiple sources in the original dialogue text, and each source is not necessarily continuous and may be distributed in different positions in the dialogue, making localization difficult. However, simply selecting only a portion of the dialogue text as input can easily result in the loss of other information, leading to lower recognition performance. Furthermore, to ensure high accuracy, currently, a separate model for predicting the content of each subheading needs to be trained, increasing the training cycle and cost.

[0005] Currently, the industry has not provided a satisfactory solution to the above problems. Summary of the Invention

[0006] This invention provides an electronic medical record generation method, an electronic device, and a storage medium to at least solve one of the above-mentioned technical problems.

[0007] In a first aspect, embodiments of the present invention provide an electronic medical record generation method, comprising: acquiring an initial dialogue text corresponding to a doctor-patient dialogue voice; for a plurality of preset case sub-topics, concatenating each case sub-topic with the initial dialogue text to determine a corresponding updated dialogue text; providing the determined updated dialogue texts to a preset topic content generation model to determine the answer content corresponding to each case sub-topic; and generating an electronic medical record based on the answer content of each case sub-topic.

[0008] In a second aspect, embodiments of the present invention provide an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the above-described method.

[0009] Thirdly, embodiments of the present invention provide a storage medium storing one or more programs including execution instructions, which can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform the steps of the method described above.

[0010] Fourthly, embodiments of the present invention also provide a computer program product, the computer program product including a computer program stored on a storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the steps of the above-described method.

[0011] The beneficial effects of the embodiments of the present invention are as follows:

[0012] The initial dialogue text corresponding to the doctor-patient conversation is concatenated with different subheadings to construct multiple updated dialogue texts. These updated dialogue texts are then used as input data for a topic content generation model, directly generating the answer content for each case subheading, thus producing an electronic medical record. Therefore, when determining the target content for each subheading, the only difference in the input data is the subheading itself. The model automatically learns and captures the importance of the subheadings and the resulting differences in outcomes, without requiring the user to locate the subheading content within the dialogue. This allows for content extraction and generation of different subheadings based on a single model, reducing the cost of generating electronic medical records while ensuring recognition accuracy. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating an example of an electronic medical record generation method according to an embodiment of the present invention is shown;

[0015] Figure 2 A flowchart illustrating an example of training a topic content generation model according to an embodiment of the present invention is shown.

[0016] Figure 3 The diagram illustrates the effect of the electronic medical record generation method based on an embodiment of the present invention in generating a structured medical record from an example doctor-patient dialogue text sample.

[0017] Figure 4 A flowchart illustrating an example of the operation process of a topic content generation model according to an embodiment of the present invention is shown.

[0018] Figure 5 A flowchart illustrating the entire lifecycle operation of an electronic medical record generation method according to an embodiment of the present invention is shown.

[0019] Figure 6 This is a schematic diagram of the structure of an embodiment of the electronic device of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0022] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, elements, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0023] In this invention, "module," "system," etc., refer to relevant entities applied to a computer, such as hardware, combinations of hardware and software, software, or software in execution. More specifically, for example, an element can be, but is not limited to, a process running on a processor, a processor, an object, an executable element, an execution thread, a program, and / or a computer. Furthermore, an application program or script running on a server, and the server itself, can also be an element. One or more elements may be in an execution process and / or thread, and elements may be localized on a single computer and / or distributed across two or more computers, and may be run on various computer-readable media. Elements can also communicate via local and / or remote processes based on signals having one or more data packets, for example, signals from data interacting with another element in a local system, a distributed system, and / or interacting with other systems via signals over a network of the Internet.

[0024] Finally, it should be noted that in this document, the terms "comprising" or "including" include not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0025] Figure 1 A flowchart illustrating an example of an electronic medical record generation method according to an embodiment of the present invention is shown. The implementing entity of this embodiment can be a client terminal device such as a mobile phone or computer, or a server device such as a cloud server; no limitation is imposed herein.

[0026] like Figure 1 As shown, in step 110, the initial dialogue text corresponding to the doctor-patient dialogue is obtained. For example, the client can call its local audio acquisition module to pick up ambient sounds in the doctor-patient dialogue scenario, such as sounds initiated by medical staff or automatically triggered by the system, and then extract the doctor-patient dialogue voice with the speaker's audio characteristics. Subsequently, the client can call its local speech recognition module to recognize the corresponding text, or send the doctor-patient dialogue voice to the server for recognition by the server's speech recognition module.

[0027] In step 120, for each of the pre-defined multiple case subtopics, each case subtopic is concatenated with the initial dialogue text to determine the corresponding updated dialogue text. Here, when preparing the model's input data, there is no need to manually locate and label the case subtopics to which different parts of the dialogue text belong in advance, which greatly reduces the preparation time and cost of the model's input data.

[0028] In step 130, the determined updated dialogue texts are provided to a preset topic content generation model to determine the answer content corresponding to each case subtopic. Here, the topic content generation model can produce different answer content according to the different case subtopics in the input updated dialogue text, and the model prediction process focuses on the case subtopics.

[0029] In step 140, an electronic medical record is generated based on the answer content of each case subtopic.

[0030] In this embodiment of the invention, the initial dialogue text corresponding to the dialogue speech is concatenated with the case subtopic text to obtain new dialogue text corresponding to each case subtopic. These new dialogue texts are then input into the topic content generation model, which automatically learns and captures the importance of the subheadings and the resulting differences, without requiring the user to locate the subheading content in the dialogue. This enables the extraction and generation of content for different subheadings based on a single model, reducing the cost of generating electronic medical records while ensuring recognition effectiveness.

[0031] Regarding the implementation details of step 110 above, firstly, voice data is collected from environmental audio data, for example, by separating silence data or environmental noise using VAD (Voice Activity Detection) technology. Then, speech recognition is performed on the collected doctor-patient dialogue to obtain the corresponding speech recognition results, and the speech recognition results are cleaned and filtered for invalid keywords to obtain the corresponding initial dialogue text. For example, invalid statements or keywords in the dialogue text, such as "hello" and "thank you," can be filtered out to reduce the proportion of invalid statements in the recognized initial dialogue text.

[0032] It should be noted that in the text corresponding to the doctor-patient dialogue, the text content of different sub-topics may not be completely independent; there may be a certain degree of correlation. For example, the content corresponding to the sub-topic "diagnosis" may be strongly correlated with the content corresponding to the sub-topic "past medical history." Therefore, when performing speech recognition, the correlation between the above-mentioned sub-topics of different cases can be considered to improve the accuracy of the recognition results.

[0033] In some examples of embodiments of the present invention, each case subtopic has a corresponding topic priority. Therefore, when predicting answer content using a topic content generation model, the updated dialogue text of the corresponding case subtopic can be provided to the topic content generation model according to topic priority order to sequentially determine the answer content corresponding to each case subtopic. Here, when determining the answer content corresponding to a non-highest topic priority, the input information of the topic content generation model also includes the answer content corresponding to a higher-level topic priority.

[0034] For example, when the topic content generation model makes predictions, it can sequentially decode different topics to obtain the answer content. For instance, when predicting answer content A for case subtopic a, it calculates based on answer content B corresponding to the higher-priority topic b and the updated dialogue text corresponding to case subtopic a. This comprehensively considers the relevance of different case subtopics, improving the accuracy of the model's predicted answer content.

[0035] In some specific implementations, the topic content generation model can adopt a Seq2Seq model to achieve a stepwise prediction process for different case subtopics, as described above.

[0036] More specifically, the topic content generation model can employ the multilingual Seq2Seq denoising autoencoder model mBART (multilingual BART). Here, mBART pre-trains a complete autoregressive Seq2Seq model. mBART is trained once for all languages, providing a set of parameters that can be fine-tuned for any language pair in supervised and unsupervised settings without requiring any task- or language-specific modifications or initialization schemes. Therefore, it supports not only generation tasks within the same language but also generation tasks between different languages.

[0037] Figure 2 A flowchart illustrating an example of training a topic content generation model according to an embodiment of the present invention is shown. Figure 3 This diagram illustrates the generation of structured medical records based on sample doctor-patient dialogue text.

[0038] like Figure 2 As shown, in step 210, a sample of the doctor-patient dialogue text is obtained.

[0039] Here, invalid statements in the dialogue, such as "hello" and "thank you," can be filtered out, reducing some invalid input, which can reduce the pressure on the model and improve its performance. Since the method in this embodiment of the invention does not require locating the content source of each subheading of the case, but instead uses the entire dialogue text as input, allowing the model to autonomously learn how to select the feature source for generating content, the input dialogue text retains as much useful information as possible while discarding useless information.

[0040] In step 220, for each individual turn text in the doctor-patient dialogue text sample, the dialogue intent corresponding to the individual turn text is identified, and the case sub-topic matching the dialogue intent is determined. The topic keywords matching the case sub-topic are extracted from the individual turn text.

[0041] In step 230, the answer content for each case subtopic is constructed based on the topic keywords corresponding to each case subtopic.

[0042] In step 240, training samples for the topic content generation model are determined based on the doctor-patient dialogue text samples, each case subtopic, and the corresponding answer content.

[0043] Reference Figure 3 For example, the one-way conversation text "Patient: My child coughs badly at night, and once he starts coughing, he can't stop" is identified as the dialogue intent "symptom disclosure," and the one-way conversation text "Doctor: Hello, how long has your child been coughing?" is identified as the dialogue intent "inquiring about symptom details," and so on. Then, the case sub-topics matching the dialogue intents are determined. For example, "symptom disclosure," "inquiring about symptom details," and "responding to symptom details" all belong to the dialogue intents related to the case sub-topics "chief complaint" and "present illness history," and thematic keywords matching the case sub-topics are extracted from the one-way conversation texts. Then, using the thematic keywords corresponding to each case sub-topic, the answer content for the corresponding case sub-topic is constructed. Thus, training samples for the topic content generation model can be constructed relatively intelligently and quickly.

[0044] Figure 4 A schematic flowchart illustrating an example of the operation process of a topic content generation model according to an embodiment of the present invention is shown. Figure 5 A flowchart illustrating the entire lifecycle operation of an electronic medical record generation method according to an embodiment of the present invention is shown.

[0045] Specifically, the subheadings and content of the medical electronic medical record are split. The cleaned dialogue is used as the Context, the subheadings as the Query, and the subheading content as the Answer. The Context and Query are concatenated to form the source text, and the target generated text is the Answer, which serves as the target text. This is used to generate the training set. As described in other sections above, the topic content generation model can use mBART to support generation tasks in both the same and different languages.

[0046] In this embodiment of the invention, the input to the topic content generation model is "dialogue text + subtitle". The two modules are separated by special characters, and the input features are constructed based on feature concatenation. This reduces both the number of training iterations and the text length of the target generated result. For different subtitles of the same electronic medical record, their inputs are largely the same, but the subtitles themselves are different. For example, for the generated result of "chief complaint", the input is "dialogue text + chief complaint", and for the generated result of "diagnosis", the input is "dialogue text + diagnosis". The only difference in the input is the subtitle, but the output target content is vastly different, thus forcing the model to learn and capture the differences and importance of the subtitles and generate content corresponding to the subtitles.

[0047] Combining the above Figure 3 In the example, the electronic medical record subheadings include "Chief Complaint," "Present Illness," and "Diagnosis." For "Chief Complaint," the model input is "Patient: My child coughs at night... Doctor: Hello, my child has a cough... Thank you, doctor! @Chief Complaint," and the target output is "Severe nighttime cough for 5 days." Similarly, for "Diagnosis," the model input is "Patient: My child coughs at night... Doctor: Hello, my child has a cough... Thank you, doctor! @Diagnosis," and the target output is "Pediatric bronchitis."

[0048] Furthermore, during the model training phase, the cleaned dialogue is used as the context, and the various subheadings in the electronic medical record are used as queries. These are concatenated and input into the generative model to obtain the corresponding Answer for each Query. mBART (multilingual BART) is a seq2seq generative pre-trained model. It supports not only generation tasks within the same language but also generation tasks between different languages. In supervised generation tasks, its training objective function is:

[0049] In supervised generation tasks, the training objective function is:

[0050]

[0051] Where X represents the result of the input text passing through the Encoder, θ represents the model parameters, and yi represents the target text sequence.

[0052] When using a topic-based content generation model for prediction, mBART takes a start identifier as input to obtain the first word of the generated result. This first word is then used as the next input to the network, and its output becomes the second word. This process continues until the output becomes the termination identifier. This decoding method allows the generation of a complete sentence. Finally, each query and its corresponding answer are populated into the case template to obtain the final electronic medical record.

[0053] In this embodiment of the invention, Context and Query features are concatenated to generate text, enabling a single model to generate answer content for different subheadings. Furthermore, by generating multi-dimensional subheading content separately and then filling it with templates, the generated medical records can be ensured to have correct formatting and diverse content. Thus, this template-based semi-supervised electronic medical record generation takes doctor-patient dialogue text as input and outputs the generated electronic medical record, without needing to determine the specific source of the content.

[0054] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps can be performed in other orders or simultaneously according to the present invention. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0055] In some embodiments, the present invention provides a non-volatile computer-readable storage medium storing one or more programs including execution instructions, which can be read and executed by an electronic device (including but not limited to a computer, server, or network device, etc.) to perform any of the above-described electronic medical record generation methods of the present invention.

[0056] In some embodiments, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform any of the above-described electronic medical record generation methods.

[0057] In some embodiments, the present invention also provides an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform an electronic medical record generation method.

[0058] Figure 6 This is a schematic diagram of the hardware structure of an electronic device for executing an electronic medical record generation method according to another embodiment of the present invention, as shown below. Figure 6 As shown, the device includes:

[0059] One or more processors 610 and memory 620, Figure 6 Take the 610 processor as an example.

[0060] The device for performing the electronic medical record generation method may further include an input device 630 and an output device 640.

[0061] The processor 610, memory 620, input device 630, and output device 640 can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.

[0062] The memory 620, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the electronic medical record generation method in this embodiment of the invention. The processor 610 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions, and modules stored in the memory 620, thereby implementing the electronic medical record generation method of the above-described method embodiment.

[0063] The memory 620 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the voice interaction device. Furthermore, the memory 620 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 620 may optionally include memory remotely located relative to the processor 610, and these remote memories can be connected to the voice interaction device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0064] Input device 630 can receive input numerical or character information and generate signals related to user settings and function control of the voice interaction device. Output device 640 may include display devices such as a display screen.

[0065] The one or more modules are stored in the memory 620, and when executed by the one or more processors 610, they execute the electronic medical record generation method in any of the above method embodiments.

[0066] The above-described product can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0067] The electronic devices of this invention exist in various forms, including but not limited to:

[0068] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include smartphones, multimedia phones, feature phones, and low-end phones.

[0069] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include: PDAs, MIDs, and UMPCs, etc.

[0070] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes audio and video players, handheld game consoles, e-book readers, as well as smart toys and portable car navigation devices.

[0071] (4) Other airborne electronic devices with data interaction capabilities, such as vehicle-mounted systems installed on vehicles.

[0072] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating electronic medical records, comprising: Obtain the initial dialogue text corresponding to the doctor-patient dialogue audio; wherein, obtaining the initial dialogue text corresponding to the doctor-patient dialogue audio includes: performing a speech recognition operation on the collected doctor-patient dialogue audio to obtain the corresponding speech recognition result; and performing data cleaning and filtering on the speech recognition result for corresponding invalid keywords to obtain the corresponding initial dialogue text; For multiple pre-defined case subtopics, each case subtopic is concatenated with the initial dialogue text to determine the corresponding updated dialogue text; The determined updated dialogue texts are provided to a preset topic content generation model to determine the answer content corresponding to each case subtopic. This includes providing the updated dialogue texts of the corresponding case subtopics to the topic content generation model according to topic priority, so as to determine the answer content corresponding to each case subtopic in turn. When determining the answer content corresponding to a non-highest topic priority, the input information of the topic content generation model also includes the answer content corresponding to a higher-level topic priority. Based on the answers to each case subtopic, a pre-defined case template is filled in to generate an electronic case.

2. The method of claim 1, wherein, The topic content generation model adopts the Seq2Seq model.

3. The method of claim 1, wherein, Before providing the determined updated dialogue texts to a preset topic content generation model to determine the answer content corresponding to each case subtopic, the method further includes: Obtain sample doctor-patient dialogue texts; For each individual turn-taking text in the doctor-patient dialogue text sample, the dialogue intent corresponding to the individual turn-taking text is identified, and the case subtopic matching the dialogue intent is determined. The topic keywords matching the case subtopic are extracted from the individual turn-taking text. Based on the keywords corresponding to each case subtopic, construct the answer content for the corresponding case subtopic; Based on the doctor-patient dialogue text samples, each case subtopic, and the corresponding answer content, training samples for the generation model for the topic content are determined.

4. The method of claim 1, wherein, The process of filling the case template with the answer content of each case subtopic to generate the corresponding electronic case includes: Get the type of the case to be generated; From multiple candidate case templates, determine the case template that matches the type of the case to be generated; The case template is populated with the answers to each case subtopic to generate the corresponding electronic case.

5. An electronic device comprising: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the method according to any one of claims 1-4.

6. A storage medium having stored thereon a computer program, characterized in that When executed by a processor, the program implements the steps of the method described in any one of claims 1-4.

Images