WO2021208444A1 - 电子病例自动生成方法、装置、设备及存储介质 - Google Patents

电子病例自动生成方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2021208444A1
WO2021208444A1 PCT/CN2020/131681 CN2020131681W WO2021208444A1 WO 2021208444 A1 WO2021208444 A1 WO 2021208444A1 CN 2020131681 W CN2020131681 W CN 2020131681W WO 2021208444 A1 WO2021208444 A1 WO 2021208444A1
Authority
WO
WIPO (PCT)
Prior art keywords
dialogue
doctor
text
medical record
unit
Prior art date
Application number
PCT/CN2020/131681
Other languages
English (en)
French (fr)
Inventor
唐蕊
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021208444A1 publication Critical patent/WO2021208444A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H15/00ICT specially adapted for medical reports, e.g. generation or transmission thereof
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35Clustering; Classification
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/10Text processing
    • G06F40/166Editing, e.g. inserting or deleting
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/279Recognition of textual entities
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/10Office automation; Time management

Definitions

  • This application relates to the field of artificial intelligence, and in particular to a method, device, equipment, and storage medium for automatically generating electronic medical records.
  • the online consultation service is carried out in the dialogue between the doctor and the patient.
  • the inventor found that after the online consultation is over, the doctor needs to manually fill in the patient’s electronic medical record based on the information obtained in the dialogue to generate a standardized electronic medical record; another way is to directly write the content of the dialogue into the electronic medical record.
  • the degree of standardization of automatically generated electronic medical records is low. Therefore, the existing electronic medical record generation technology has the problem that it is difficult to extract the key consultation content based on the content of the consultation dialogue to automatically generate a standardized electronic medical record.
  • the main purpose of this application is to solve the technical problem that the prior art is difficult to automatically generate a standardized electronic medical record.
  • the first aspect of this application provides a method for automatically generating electronic medical records, including:
  • the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • each patient dialogue unit and each doctor dialogue unit sequentially archive each patient dialogue unit and each doctor dialogue unit to generate a corresponding initial electronic medical record
  • the second aspect of the present application provides an automatic generation device for an electronic medical record, including:
  • the construction module is used to obtain the dialogue text in the online medical inquiry scene, and construct a plurality of dialogue text units according to the dialogue text, wherein the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • the classification module is configured to use a preset patient dialogue classification model and a preset doctor dialogue classification model to classify each patient dialogue unit and each doctor dialogue unit respectively to obtain the corresponding medical record item category;
  • a generating module configured to sequentially archive each patient dialogue unit and each doctor dialogue unit according to the category of the medical record item, and generate a corresponding initial electronic medical record
  • the standardization module is used to standardize the initial electronic case to obtain the corresponding electronic case.
  • the third aspect of the present application provides an electronic medical record automatic generation device, including: a memory and at least one processor, the memory stores instructions; the at least one processor calls the instructions in the memory to make The electronic case automatic generation device executes the steps of the electronic case automatic generation method as follows:
  • the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • each patient dialogue unit and each doctor dialogue unit sequentially archive each patient dialogue unit and each doctor dialogue unit to generate a corresponding initial electronic medical record
  • the fourth aspect of the present application provides a computer-readable storage medium having instructions stored in the computer-readable storage medium, which when run on a computer, cause the computer to execute the method for automatically generating an electronic medical record as shown below :
  • the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • each patient dialogue unit and each doctor dialogue unit sequentially archive each patient dialogue unit and each doctor dialogue unit to generate a corresponding initial electronic medical record
  • the patient dialogue unit and the doctor dialogue text unit are constructed for the dialogue text between the patient and the doctor in the online consultation scenario; then the patient dialogue unit and the doctor dialogue text unit are respectively input into the corresponding patient text Text classification is performed in the classification model and the doctor text classification model; then the words of the classified patients and doctors obtained through the classification model are archived in the medical record items; finally the classified patient texts are standardized to generate the electronic planning specification Medical record.
  • FIG. 1 is a schematic diagram of the first embodiment of the method for automatically generating an electronic medical record in an embodiment of the application
  • FIG. 2 is a schematic diagram of a second embodiment of the method for automatically generating an electronic medical record in an embodiment of the application
  • FIG. 3 is a schematic diagram of a third embodiment of the method for automatically generating an electronic medical record in an embodiment of the application;
  • FIG. 4 is a schematic diagram of a fourth embodiment of the method for automatically generating an electronic medical record in an embodiment of the application;
  • FIG. 5 is a schematic diagram of an embodiment of an electronic medical record automatic generation device in an embodiment of the application.
  • Fig. 6 is a schematic diagram of another embodiment of the automatic generation device of an electronic medical record in an embodiment of the application.
  • Fig. 7 is a schematic diagram of an embodiment of an automatic generation device for an electronic medical record in an embodiment of the application.
  • the embodiments of the present application provide a method, device, equipment, and storage medium for automatically generating an electronic medical record.
  • the dialog text in an online consultation scene is acquired and multiple dialog text units are constructed.
  • the dialog text unit includes a patient dialog unit.
  • Dialogue text unit with the doctor use the preset patient dialogue classification model and the preset doctor dialogue classification model to classify each patient dialogue unit and each doctor dialogue unit to obtain the corresponding medical record item category; according to the medical record item category, each patient is classified in turn
  • the dialogue unit and each doctor dialogue unit perform filing operations to generate the corresponding initial electronic case; standardize the initial electronic case to obtain the corresponding electronic case.
  • This application also relates to blockchain technology, and the dialogue text is stored in the blockchain. This application realizes the automatic generation of electronic medical records after the patient's online consultation, and improves the standardization of the generation of electronic medical records.
  • the first embodiment of the method for automatically generating an electronic medical record in the embodiment of the present application includes:
  • the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • the execution subject of this application may be an electronic medical record automatic generation device, or a terminal or a server, which is not specifically limited here.
  • the embodiment of the present application takes the server as the execution subject as an example for description. It should be emphasized that, in order to further ensure the privacy and security of the above-mentioned dialogue text, the above-mentioned dialogue text can also be stored in a node of a blockchain.
  • the online consultation is conducted as a dialogue between the patient and the doctor.
  • the dialogue is divided into dialogue texts, where each sentence of the patient’s speech in the dialogue text is regarded as a patient Dialogue text, each sentence of the doctor's speech is regarded as a doctor's dialogue text.
  • the dialogue text unit refers to the smallest text unit extracted from the dialogue text and used for subsequent model classification.
  • the patient dialogue unit can be: each sentence of the patient dialogue text, each sentence of the patient dialogue text and the doctor dialogue context; and the doctor dialogue
  • the text unit can be: each sentence of the doctor conversation text, and each sentence of the doctor conversation text splicing the patient conversation context.
  • the medical record item category refers to the filled-in items appearing in the electronic medical record, including the chief complaint, current medical history, past history, family history, etc.; for the patient dialogue classification model, the patient dialogue unit is used as input, and the patient dialogue unit medical record items The category is the output; for the doctor dialogue classification model, the doctor dialogue unit is the input, and the medical record item category of the doctor dialogue unit is the output.
  • a patient dialogue unit or a doctor dialogue unit can be classified into one or more medical record item categories.
  • patient dialogue unit A records the family history of the patient who has been diagnosed with disease a
  • patient dialogue unit B records: the patient’s past history has disease b
  • the patient’s chief complaint is disease c.
  • each patient dialogue unit and doctor dialogue unit After obtaining each patient dialogue unit and doctor dialogue unit, it also includes:
  • other types of items refer to items that have nothing to do with the patient's disease, which are specifically manifested in that there are no corresponding filling items in the electronic medical record; or the specific medical record item category of the dialogue text unit cannot be identified.
  • the dialogue text units that do not need to be written into the electronic medical record or cannot be recognized are all classified into other item categories and can be directly eliminated.
  • each patient dialogue unit and each doctor dialogue unit sequentially archive each patient dialogue unit and each doctor dialogue unit to generate a corresponding initial electronic medical record
  • both the patient dialogue unit and the doctor dialogue unit are classified, and the corresponding medical record item category is labeled; the patient dialogue unit and the doctor dialogue unit are written into the corresponding columns of the electronic medical record initialization template according to the label annotation;
  • the standardized processing of the initial electronic medical record mainly includes the following two parts:
  • the word frequency is relatively high in general scenarios. Therefore, the word frequency of each word in the initial electronic medical record in the general scenario is counted, and the words with high word frequency can be screened out as non-medical vocabulary.
  • the patient dialogue unit and the doctor dialogue text unit construct the patient dialogue unit and the doctor dialogue text unit; then, input the patient dialogue unit and the doctor dialogue text unit into the corresponding patient text respectively Text classification is performed in the classification model and the doctor text classification model; next, the patient’s words and doctors’ words after classification obtained through the classification model are archived for medical record items; finally the classified patient texts are standardized to generate a planning model Electronic medical records.
  • the second embodiment of the method for automatically generating an electronic medical record in the embodiment of the present application includes:
  • the doctor as a professional, generally speaks in a targeted manner, and the obtained doctor dialogue text is likely to contain medical information. Therefore, each sentence of the doctor dialogue text can be directly used as a doctor dialogue text unit.
  • the dialogue text T is: [Doctor 1, disease 2, medical 3, disease 4, disease 5, medical 6, medical 7, disease 8, medical 9, disease 10], where " ⁇ ” is the doctor dialog text , "Disease” patient dialogue text, 1-10 is the order of speaking. Therefore, from the dialogue text, 5 doctor text units can be extracted: ⁇ 1 ⁇ , ⁇ 3 ⁇ , ⁇ 6 ⁇ , ⁇ 7 ⁇ , ⁇ 9 ⁇ .
  • the doctor's dialogue context can be used to assist the classification of the patient's dialogue text to determine the medical information contained in the patient's dialogue text.
  • the matching anchor point is used as the classification target, which is used in the subsequent dialogue text classification model to classify the electronic medical record items of the patient dialogue text; and the doctor dialogue context is used to assist the classification of the matching anchor point, and the doctor dialogue context is used to infer and evaluate the matching anchor Whether the classification result of the point is correct, the further the distance between the doctor’s dialogue text and the speaking sequence of the matching anchor point, the worse the effectiveness of its speculation and evaluation.
  • a doctor conversation text before and/or after the matching anchor point is selected as the doctor conversation context.
  • the selection method of the doctor's dialogue context is as follows:
  • doctor conversation texts before and after the matching anchor point use one or more doctor conversation texts before and after the matching anchor point as the doctor conversation context;
  • doctor conversation texts before or after the matching anchor point If there are one or more doctor conversation texts before or after the matching anchor point, one or more doctor conversation texts before or after the matching anchor point are used as the doctor conversation context.
  • the doctor conversation text and the matching anchor point in the doctor conversation context have a sequence of speaking, and the doctor conversation context needs to be spliced to the matching anchor point in this sequence.
  • each patient dialogue unit and each doctor dialogue unit sequentially archive each patient dialogue unit and each doctor dialogue unit to generate a corresponding initial electronic medical record
  • each doctor dialogue text is used as the doctor dialogue unit for medical record items
  • the classification of categories through the context of the doctor's dialogue, assists in the classification of the medical record item category of each patient's dialogue text, and the classification accuracy is higher.
  • the third embodiment of the method for automatically generating an electronic medical record in the embodiment of the present application includes:
  • 301 Acquire a dialogue text in an online consultation scene, and construct a plurality of dialogue text units according to the dialogue text, wherein the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • each patient dialogue unit and each doctor dialogue unit sequentially archive each patient dialogue unit and each doctor dialogue unit to generate a corresponding initial electronic medical record
  • each sentence of dialogue text in the initial electronic medical record is segmented to obtain multiple segments.
  • HMM Hidden Markov Model
  • Trie word search tree
  • index tree index tree
  • HMM can be used. It is hoped that the word segmentation can be processed by indexing, CRF (Conditional Random Field) model, SVM (support vector machines, support vector machine), and deep learning.
  • the preset word frequency prior knowledge refers to the word frequency of each vocabulary in a general scenario.
  • crawler technology can be used to crawl data information in general fields, including life journals, publications in various fields, newspapers, news releases, blogs, etc.; then, statistics of vocabulary word frequencies appearing in data information; construct prior knowledge of word frequencies through word frequencies.
  • Non-medical participle refers to the participle that does not contain medical information; medical participle refers to the participle that may contain medical information.
  • Vocabulary is screened out as non-medical participles, such as "Hello”, “Excuse me”, “Ask”, “I”. These are more common in consultation scenes and in daily life, but there is no substantial medical information. , So it can be screened out directly.
  • the unscreened word segmentation after it is subsequently classified by the dialogue text classification model, it can be further screened out as "non-standard medical word segmentation" through other item categories.
  • non-standard medical word segmentation traverse a preset standard medical vocabulary semantic database, and filter the standard medical vocabulary matching the non-standard medical word segmentation from the standard medical vocabulary semantic database;
  • the standard medical vocabulary semantic database contains each standard medical vocabulary.
  • classification methods can be used: International Classification of Diseases (ICD), logical observation identifier identifiers names and codes, LOINC), medical system nomenclature of medicine--clinical terms (systematized nomenclature of medicine--clinical terms, SNOMED CT).
  • the non-standard medical participles are analyzed and the semantic features of the standard medical participles in the standard medical vocabulary semantic database are analyzed; then the matching scores between the non-standard medical participles and the standard medical participles are calculated through semantic feature matching; then the matching score is selected The highest marked medical word segmentation is sufficient.
  • the patient describes "belly pain” as a non-standard medical participle, and the standard medical vocabulary is "abdominal pain”.
  • the standard medical vocabulary is "abdominal pain”.
  • the non-standard medical vocabulary is replaced by annotated medical vocabulary, and then re-spliced and displayed on the electronic medical record to obtain a standardized electronic medical record.
  • non-medical vocabulary is eliminated, and non-standard medical vocabulary is replaced with standard medical vocabulary, so that the electronic medical record is more standardized.
  • the fourth embodiment of the method for automatically generating an electronic medical record in the embodiments of the present application includes:
  • the dialog text unit can be automatically or manually labeled to determine the medical record item category of the dialog text unit: chief complaint, current medical history, past history, or family history, etc., while the dialog text unit without medical information is labeled other The item category is fine.
  • the pre-training model may be BERT (Bidirectional Encoder Representations from Transformers, a bidirectional language model based on converters), Roberta (Robustly optimized Bidirectional Encoder Representations from Transformers approach, a robust optimization bidirectional language method model based on converters), Albert (A Lite Bidirectional Encoder Representations from Transformers, a lightweight bidirectional language model based on converters).
  • the segment mark of the model is converted to "A, B, A" to represent each sentence of the dialogue text in the patient dialogue unit.
  • the patient dialogue unit ⁇ Doctora, Diseaseb, Medicalc ⁇ are respectively expressed as [0, 1, 2], and input into the pre-training model to correspond to "A, B, A" in the model.
  • predicting the medical record item category refers to the output of the dialogue text unit after the pre-training model, and the predicted probability of each medical record item category.
  • the loss function is used to calculate the prediction accuracy of the pre-training model.
  • the loss function can use cross entropy, KL divergence (Kullback-Leibler divergence), negative log-likelihood (negative log-likelihood) function, etc.
  • cross entropy can be used as the loss function of the pre-training model, as shown below:
  • N is the number of dialogue text units
  • y (i) represents the prediction result of the i-th dialogue text unit (that is, the prediction is correct or the prediction is wrong, the prediction is correct is 1, the prediction error is 0)
  • Los represents the loss value.
  • the historical dialogue text unit is constructed as a training sample through the historical dialogue text, and the pre-training model is used to train the classification of the dialogue text unit to make the electronic medical record Realize automatic generation.
  • An embodiment of the automatic generation device for an electronic case in the embodiment of the application includes:
  • the construction module 501 is used to obtain the dialogue text in the online medical inquiry scene, and construct a plurality of dialogue text units according to the dialogue text, wherein the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • the classification module 502 is configured to use a preset patient dialogue classification model and a preset doctor dialogue classification model to classify each patient dialogue unit and each doctor dialogue unit to obtain the corresponding medical record item category;
  • the generating module 503 is configured to sequentially archive each patient dialogue unit and each doctor dialogue unit according to the medical record item category, and generate corresponding initial electronic medical records;
  • the standardization module 504 is used to perform standardization processing on the initial electronic case to obtain a corresponding electronic case.
  • the patient dialogue unit and the doctor dialogue text unit construct the patient dialogue unit and the doctor dialogue text unit; then, input the patient dialogue unit and the doctor dialogue text unit into the corresponding patient text respectively Text classification is performed in the classification model and the doctor text classification model; next, the patient’s words and doctor’s words after classification obtained through the classification model are archived for medical record items; finally the classified patient text is standardized to generate a planning model Electronic medical records.
  • FIG. 6 another embodiment of the apparatus for automatically generating an electronic medical record in the embodiment of the present application includes:
  • the construction module 501 is used to obtain the dialogue text in the online medical inquiry scene, and construct a plurality of dialogue text units according to the dialogue text, wherein the dialogue text unit includes a patient dialogue unit and a doctor dialogue text unit;
  • the classification module 502 is configured to use a preset patient dialogue classification model and a preset doctor dialogue classification model to classify each patient dialogue unit and each doctor dialogue unit to obtain the corresponding medical record item category;
  • the generating module 503 is configured to sequentially archive each patient dialogue unit and each doctor dialogue unit according to the medical record item category, and generate corresponding initial electronic medical records;
  • the standardization module 504 is used to perform standardization processing on the initial electronic case to obtain a corresponding electronic case.
  • the dialog text includes one or more doctor dialog texts and patient dialog texts
  • the construction module 501 includes:
  • the first selection unit 5011 is configured to select all doctor conversation texts from the conversation text, and use each doctor conversation text as a doctor conversation text unit;
  • the second selecting unit 5012 is configured to sequentially select a patient dialogue text from the dialogue text as a matching anchor point, and select one or more doctor dialogue texts before and/or after the matching anchor point as a doctor dialogue Context
  • the splicing unit 5013 is used to splice the doctor dialogue context and the matching anchor point to obtain a corresponding patient dialogue unit.
  • the second selecting unit 5012 is configured to:
  • doctor conversation texts before and after the matching anchor point use one or more doctor conversation texts before and after the matching anchor point as the doctor conversation context;
  • doctor conversation texts before or after the matching anchor point If there are one or more doctor conversation texts before or after the matching anchor point, one or more doctor conversation texts before or after the matching anchor point are used as the doctor conversation context.
  • the standardization module 504 includes:
  • the word segmentation unit 5041 is configured to perform word segmentation processing on the dialogue text in the initial electronic medical record to obtain multiple word segmentation;
  • the screening unit 5042 is used for screening non-medical segmentation from each of the segmentation according to the prior knowledge of preset word frequency, and taking the unscreened segmentation as non-standard medical segmentation;
  • the matching unit 5043 is configured to traverse a preset standard medical vocabulary semantic database according to the non-standard medical word segmentation, and filter the standard medical vocabulary matching the non-standard medical word segmentation from the standard medical vocabulary semantic database;
  • the replacement unit 5044 is configured to replace the corresponding non-standard medical segmentation word with the standard medical vocabulary to obtain an electronic medical record.
  • the electronic medical record automatic generation device further includes a model generation module 505, and the model generation module 505 includes:
  • the construction unit 5051 is used to obtain historical dialogue texts, and construct multiple historical dialogue text units according to the historical dialogue texts, wherein the historical dialogue text units include a historical patient dialogue unit and a historical doctor dialogue text unit;
  • the labeling unit 5052 is used to label the medical record item category of the historical dialogue text unit to obtain corresponding labeling information
  • the training unit 5053 is configured to use a preset pre-training model to train the historical dialogue text unit, and output the corresponding predicted medical record item category;
  • the calculation unit 5054 is configured to calculate the loss function of the pre-training model according to the annotation information and the predicted medical record item category;
  • the iterative unit 5055 is configured to iterate the pre-training model until the loss function is less than a preset threshold to obtain a dialog text classification model, where the text classification model is a patient dialog classification model or a doctor dialog classification model.
  • the method before the generating module 503, the method further includes:
  • the elimination module 506 is used to select other item categories in the medical record item category, and eliminate patient dialogue units and doctor dialogue units in the other item categories.
  • the generation process of the patient dialogue unit and the doctor dialogue unit is introduced in detail, and the corresponding dialogue text unit is constructed by using the characteristics of the doctor role and the patient role, and its classification accuracy is higher; and through the standardized processing of electronic medical records, The non-medical vocabulary is eliminated, and the non-standard medical vocabulary is replaced with the standard medical vocabulary to make the electronic medical record more standardized; then the generation process of the patient text classification model and the doctor text classification model is introduced in detail, and the historical dialogue text unit is constructed through the historical dialogue text.
  • the pre-training model is used to train the classification of the dialogue text unit, so that the electronic medical record can be automatically generated.
  • FIG. 7 is a schematic structural diagram of an automatic generation device for an electronic medical record provided by an embodiment of the present application.
  • the automatic generation device for an electronic medical record 700 may have relatively large differences due to different configurations or performance, and may include one or more processors (central Processing units, CPU) 710 (for example, one or more processors) and memory 720, and one or more storage media 730 for storing application programs 733 or data 732 (for example, one or one storage device with a large amount of storage).
  • the memory 720 and the storage medium 730 may be short-term storage or persistent storage.
  • the program stored in the storage medium 730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the electronic medical record automatic generation device 700.
  • the processor 710 may be configured to communicate with the storage medium 730, and execute a series of instruction operations in the storage medium 730 on the electronic medical record automatic generation device 700.
  • the electronic medical record automatic generation device 700 may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input and output interfaces 760, and/or one or more operating systems 731, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc.
  • operating systems 731 such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc.
  • This application also provides an electronic medical record automatic generation device, which includes a memory and a processor.
  • the memory stores computer-readable instructions.
  • the processor executes each of the above. The steps of the automatic generation method of the electronic medical record in the embodiment.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium, and the computer-readable storage medium may also be a volatile computer-readable storage medium.
  • the computer-readable storage medium stores instructions, and when the instructions run on a computer, the computer executes the steps of the method for automatically generating an electronic medical record.
  • the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • the technical solution of the present application essentially or the part that contributes to the existing technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , Including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (read-only memory, ROM), random access memory (random access memory, RAM), magnetic disks or optical disks and other media that can store program codes. .
  • the blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Data Mining & Analysis (AREA)
  • Strategic Management (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Primary Health Care (AREA)
  • Epidemiology (AREA)
  • Databases & Information Systems (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

一种电子病例自动生成方法、装置、设备及存储介质,涉及人工智能领域。该方法包括:获取线上问诊场景下的对话文本,并构造多个对话文本单元,其中,对话文本单元包括病人对话单元和医生对话文本单元(101);采用预置病人对话分类模型、预置医生对话分类模型分别对各病人对话单元和各医生对话单元进行分类,得到对应的病历项目类别(102);根据病历项目类别,依次将各病人对话单元和各医生对话单元进行归档操作,生成对应的初始电子病例(103);对初始电子病例进行标准化处理,得到对应的电子病例。本申请还涉及区块链技术,所述对话文本存储于区块链中(104)。实现了病人在线问诊后自动生成电子病历,提升了生成电子病历的规范化。

Description

电子病例自动生成方法、装置、设备及存储介质
本申请要求于2020年9月7日提交中国专利局、申请号为202010926413.8、发明名称为“电子病例自动生成方法、装置、设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在申请中。
技术领域
本申请涉及人工智能领域,尤其涉及一种电子病例自动生成方法及、装置、设备及存储介质。
背景技术
近些年来,随着互联网技术的快速发展,在线问诊服务变得越来越热门,在线问诊的需求也在不断攀升。通过在线问诊,人们如果感到身体不舒服,可以随时随地、方便地通过网络直接和专业的医生进行就诊沟通,然后根据医生的建议再采取进一步的就诊方案。在沟通过程中,通过将问诊内容进行记录,整理成个人的病历,在后续继续线上问诊,或者线下医院问诊时,医生可以根据个人的过往病历,辅助诊相关疾病。
而在线问诊服务是在医生和病人的对话中展开。发明人发现,在线问诊结束后,医生需要根据在对话中获取的信息,手动填写该病人的电子病历,生成规范化的电子病历;另一种方式是直接将涉及对话内容写进电子病历中,自动生成的电子病历规范化程度低。故现有电子病历生成技术存在难以根据问诊对话内容提取关键问诊内容,以自动生成规范化电子病历的问题。
发明内容
本申请的主要目的在于解决现有技术难以自动生成规范化电子病历的技术问题。
本申请第一方面提供了一种电子病例自动生成方法,包括:
获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
对所述初始电子病例进行标准化处理,得到对应的电子病例。
本申请第二方面提供了一种电子病例自动生成装置,包括:
构造模块,用于获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
分类模块,用于采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
生成模块,用于根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
标准化模块,用于对所述初始电子病例进行标准化处理,得到对应的电子病例。
本申请第三方面提供了一种电子病例自动生成设备,包括:存储器和至少一个处理器,所述存储器中存储有指令;所述至少一个处理器调用所述存储器中的所述指令,以使得所述电子病例自动生成设备执行如下所示的电子病例自动生成方法的步骤:
获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
对所述初始电子病例进行标准化处理,得到对应的电子病例。
本申请的第四方面提供了一种计算机可读存储介质,所述计算机可读存储介质中存储有指令,当其在计算机上运行时,使得计算机执行如下所示的电子病例自动生成方法的方法:
获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
对所述初始电子病例进行标准化处理,得到对应的电子病例。
本申请提供的技术方案中,首先针对在线问诊场景下病人和医生的对话文本,进行病人对话单元、医生对话文本单元构造;然后将病人对话单元、医生对话文本单元分别输入到对应的病人文本分类模型、医生文本分类模型中进行文本分类;接着将通过分类模型得到的分类后的病人的话和医生的话进行病历项目的归档;最后对归类好的病人文本进行标准化,以生成规划范的电子病历。
附图说明
图1为本申请实施例中电子病例自动生成方法的第一个实施例示意图;
图2为本申请实施例中电子病例自动生成方法的第二个实施例示意图;
图3为本申请实施例中电子病例自动生成方法的第三个实施例示意图;
图4为本申请实施例中电子病例自动生成方法的第四个实施例示意图;
图5为本申请实施例中电子病例自动生成装置的一个实施例示意图;
图6为本申请实施例中电子病例自动生成装置的另一个实施例示意图;
图7为本申请实施例中电子病例自动生成设备的一个实施例示意图。
具体实施方式
本申请实施例提供了一种电子病例自动生成方法、装置、设备及存储介质,通过获取线上问诊场景下的对话文本,并构造多个对话文本单元,其中,对话文本单元包括病人对话单元和医生对话文本单元;采用预置病人对话分类模型、预置医生对话分类模型分别对各病人对话单元和各医生对话单元进行分类,得到对应的病历项目类别;根据病历项目类别,依次将各病人对话单元和各医生对话单元进行归档操作,生成对应的初始电子病例;对初始电子病例进行标准化处理,得到对应的电子病例。本申请还涉及区块链技术,所述对话文本存储于区块链中。本申请实现了病人在线问诊后自动生成电子病历,提升了生成电子病历的规范化。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”、“第三”、“第四”等(如果存在)是用于区别类似的对象,而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换,以便这里描述的实施例能够以除了在这里图示或描述的内容以外的顺序实施。此外,术语“包括”或“具有”及其任何变形,意图在于覆盖不排他的包含,例如,包含了一系列步骤或单元的过程、方法、系统、产品或设备不 必限于清楚地列出的那些步骤或单元,而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
为便于理解,下面对本申请实施例的具体流程进行描述,请参阅图1,本申请实施例中电子病例自动生成方法的第一个实施例包括:
101、获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
可以理解的是,本申请的执行主体可以为电子病例自动生成装置,还可以是终端或者服务器,具体此处不做限定。本申请实施例以服务器为执行主体为例进行说明。需要强调的是,为进一步保证上述对话文本的私密和安全性,上述对话文本还可以存储于一区块链的节点中。
本实施例中,线上问诊以病人与医生的对话展开,从病人开始提问到最后问诊结束,将对话分割出来,即为对话文本,其中,对话文本中病人的每一句发言作为一个病人对话文本,医生的每一句发言作为一个医生对话文本。对话文本单元指从对话文本中提取出来的,用于后续模型分类的最小文本单元,其中,对于病人对话单元可以为:每一句病人对话文本、每一句病人对话文本拼接医生对话上下文;而医生对话文本单元可以为:每一句医生对话文本、每一句医生对话文本拼接病人对话上下文。
102、采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
本实施例中,病历项目类别指电子病历中出现的填写的项目,包括主诉、现病史、既往史、家族史等;对于病人对话分类模型,以病人对话单元为输入,病人对话单元的病历项目类别为输出;对于医生对话分类模型,以医生对话单元为输入,医生对话单元的病历项目类别为输出。其中,一个病人对话单元或医生对话单元可以归入一个或多个病历项目类别。
比如,病人对话单元A记录有病人的家族史曾确诊疾病a、病人对话单元B记录有:病人的既往史有疾病b,病人的主诉为疾病c。将病人对话单元A与病人对话单元B输入病人对话分类模型中,即可得到病人对话单元A的病历项目类别为“家族史”,病人对话单元B的病历项目类别为“既往史”和“主诉”。
具体的,在得到每一个病人对话单元与医生对话单元后,还包括:
选取所述病历项目类别中的其他项目类别,并剔除所述其他项目类别中的病人对话单元和医生对话单元;
本实施例中,其他类别项目指的是与病人疾病无关的项目,具体表现为在电子病历中,没有相应的填写项目;或者是无法识别出该对话文本单元的具体病历项目类别。对于无需写入电子病历,或无法识别的对话文本单元均归为其他项目类别,直接剔除即可。
103、根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
本实施例中,病人对话单元和医生对话单元均进行分类,以对应的病历项目类别进行标签标注;根据标签标注,将病人对话单元和医生对话单元写入电子病历初始化模板的对应栏目中;
104、对所述初始电子病例进行标准化处理,得到对应的电子病例。
本实施例中,初始电子病历的标准化处理主要包括以下两部分:
由于病人和医生输入的内容,其语言风格不统一,较为杂乱,所以将初始电子病历中的非标准医学词汇替换为标准医疗词汇;
由于问诊过程中,病人和医生的输入内容较为口语化,包含除医学信息外的用词,故需将不重要的词汇剔除,包括停用词、感叹词等没有医学信息含义的词汇。
本实施例中,对于第一部分中的非标准医疗词汇,使用预先构建好的标准医疗词汇语义库,筛选对应的标注医疗词汇进行替换即可;对于第二部分中的非医疗词汇,由于具有口语化的特点,故其词频在一般场景下均较高,故通过统计初始电子病历中各分词在普通场景下的词频,并将词频高的用词作为非医疗词汇进行筛除即可。
本申请实施例中,首先,针对在线问诊场景下病人和医生的对话文本,进行病人对话单元、医生对话文本单元构造;然后,将病人对话单元、医生对话文本单元分别输入到对应的病人文本分类模型、医生文本分类模型中进行文本分类;接下来,将通过分类模型得到的分类后的病人的话和医生的话进行病历项目的归档;最后对归类好的病人文本进行标准化,以生成规划范的电子病历。
请参阅图2,本申请实施例中电子病例自动生成方法的第二个实施例包括:
201、获取线上问诊场景下的对话文本;
202、从所述对话文本中选取全部医生对话文本,并将所述各医生对话文本作为医生对话文本单元;
本实施例中,医生作为专业人士,一般具有针对性地进行发言,得到的医生对话文本中,包含医学信息的可能性较高,故可直接将每一句医生对话文本作为一个医生对话文本单元。
具体的,若对话文本T为:[医1,病2,医3,病4,病5,医6,医7,病8,医9,病10],其中,“医”为医生对话文本,“病”病人对话文本,1-10为发言顺序。故从该对话文本中,可以提取到5个医生文本单元:{{医1},{医3},{医6},{医7},{医9}}。
203、依次从所述对话文本中选取一个病人对话文本作为匹配锚点,并选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文;
本实施例中,由于病人并非医学领域专业人士,故可通过医生对话上下文辅助病人对话文本进行分类,确定病人对话文本中包含的医学信息。匹配锚点作为分类目标,用于后续的对话文本分类模型,对病人对话文本进行电子病历项目的分类;而医生对话上下文用于对匹配锚点进行辅助分类,使用医生对话上下文推测及评估匹配锚点的分类结果是否正确,其中医生对话文本与匹配锚点的发言顺序距离越远,其推测及评估有效性越差。
优选地,选取匹配锚点之前和/或之后的一个医生对话文本作为医生对话上下文。
具体的,医生对话上下文的选取方法如下所示:
判断在所述匹配锚点之前是否存在一个或多个医生对话文本以及在所述匹配锚点之后是否存在一个或多个医生对话文本;
若在所述匹配锚点之前和之后均存在一个或多个医生对话文本,则将所述匹配锚点之前和之后的一个或多个医生对话文本作为医生对话上下文;
若在所述匹配锚点之前或之后存在一个或多个医生对话文本,则将所述匹配锚点之前或之后的一个或多个医生对话文本作为医生对话上下文。
比如,对于对话文本T,依次将“病2”、“病4”、“病5”、“病8”、“病10”作为匹配锚点,对于“病2”,可以将“医1”、“医3”作为医生对话上下文;对于“病4”,可以将“医3”、“医6”作为医生对话上下文;对于“病10”,由于“病10”之后没有医生对话文本,故只需将“医9”作为医生对话上下文即可,以此类推得到“病2”、“病4”、“病5”、“病8”、“病10”的医生对话上下文。
204、将所述医生对话上下文和所述匹配锚点进行拼接,得到对应的病人对话单元;
本实施例中,医生对话上下文中的医生对话文本与匹配锚点具有发言的前后顺序,需以该前后顺序将医生对话上下文拼接到匹配锚点上。
比如,对于对话文本T,可以将“医1”、“医3”与“病2”进行拼接,将“医3”、“医6” 与“病4”进行拼接,将“医3”、“医6”与“病5”进行拼接,将“医7”、“医9”与“病8”进行拼接,将“医9”与“病10”进行拼接,得到对应的病人对话单元:{{医1,病2,医3},{医3,病4,医6},{医3,病5,医6},{医7,病8,医9},{医9,病10}}。
205、采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
206、根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
207、对所述初始电子病例进行标准化处理,得到对应的电子病例。
本申请实施例中,详细介绍了病人对话单元和医生对话单元的生成过程,利用医生角色和病人角色的特性构造对应的对话文本单元,其中,将每一个医生对话文本作为医生对话单元进行病历项目类别的分类,通过医生对话上下文对每一个病人对话文本的病历项目类别进行辅助分类,其分类准确性更高。
请参阅图3,本申请实施例中电子病例自动生成方法的第三个实施例包括:
301、获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
302、采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
303、根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
304、对所述初始电子病历中的对话文本进行分词处理,得到多个分词;
本实施例中,对初始电子病历中每一句对话文本进行分词,得到多个分词,具体可使用HMM(Hidden Markov Model,隐含马尔柯夫模型)、Trie(单词查找树)、索引树、哈希索引、CRF(Conditional Random Field,条件随机场)模型、SVM(support vector machines,支持向量机)、深度学习等方式进行分词处理。
305、根据预置词频先验知识,从所述各分词中筛除非医学分词,并将未筛除的分词作为非标准医学分词;
本实施例中,预置词频先验知识指:通用场景下各词汇的词频。具体可通过爬虫技术爬取通用领域下的资料信息,包括生活期刊、各领域刊物、报纸、新闻稿、博客等;然后统计资料信息出现的词汇词频;通过词频构建词频先验知识。非医学分词指未包含医学信息的分词;医学分词指可能包含有医学信息的分词。
进一步的,对于生活中常用的词汇,包括语气助词、停用词、敬语等,在普通场景下,其出现频率一般较高,故通过先验知识,将初始电子病历中,词频较高的词汇作为非医学分词进行筛除,比如“你好”、“请问”、“请教”、“本人”,这些在问诊场景中较为常见,在日常中亦较为常见,但并无实质的医学信息,故可直接筛除。
本实施例中,对于未筛除的分词,在后续经过对话文本分类模型分类后,可以通过其他项目类别,对判定为“非标准医学分词”作进一步的筛除。
306、根据所述非标准医疗分词,遍历预置标准医疗词汇语义库,从所述标准医疗词汇语义库中筛选与所述非标准医疗分词相匹配的标准医疗词汇;
本实施例中,标准医疗词汇语义库包含各标准医疗词汇,具体可使用一下分类方法:国际疾病分类(International Classification of Diseases,ICD)、观测指标标识符逻辑命名与编码系统(logical observation identifiers names and codes,LOINC)、医学系统命名法—临床术语(systematized nomenclature of medicine--clinical terms,SNOMED CT)。
本实施例中,分析非标准医疗分词,与标准医疗词汇语义库中标准医疗分词的语义特征;然后通过语义特征匹配,计算出非标准医疗分词与各标准医疗分词的匹配得分;接着选择匹配得分最高的标注医疗分词即可。
比如病人描述“肚子痛”为非标准医疗分词,标准医疗词汇为“腹痛”,通过前后医生对话文本,还可能可以推断出具体“腹痛”位置,如“右侧肩背部的射痛”、“左腰部射痛”、“阴部射痛”等。
307、通过所述标准医疗词汇替换对应的非标准医疗分词,得到电子病历。
本实施例中,将标注医疗词汇替代非标准医疗词汇,然后重新拼接,显示在电子病历上,即可得到规范化的电子病历。
本申请实施例中,通过电子病历的标准化处理,将非医学词汇剔除,将非标准医疗词汇以标准医疗词汇进行代替,使得电子病历更规范。
请参阅图4,本申请实施例中电子病例自动生成方法的第四个实施例包括:
401、获取历史对话文本,并根据所述历史对话文本,构造多个历史对话文本单元,其中,所述历史对话文本单元包括历史病人对话单元和历史医生对话文本单元;
对所述历史对话文本单元的病历项目类别进行标注,得到对应标注信息;
本实施例中,可通过自动标注或者人工标注对话文本单元,确定对话文本单元的病历项目类别为主诉、现病史、既往史或家族史等,而未带有医学信息的对话文本单元,标注其他项目类别即可。
402、采用预置预训练模型对所述历史对话文本单元进行训练,输出对应的预测病历项目类别;
本实施例中,预训练模型可以为BERT(Bidirectional Encoder Representations from Transformers,基于转换器的双向语言模型)、Roberta(Robustly optimized Bidirectional Encoder Representations from Transformers approach,基于转换器的稳健优化双向语言方法模型)、Albert(A Lite Bidirectional Encoder Representations from Transformers,基于转换器的轻量双向语言模型)。
其中,当使用预训练模型对病人对话单元进行训练时,将模型的segement标记转换为“A、B、A”,以表示病人对话单元中的每一句对话文本。将病人对话单元{医a,病b,医c}分别表示为[0,1,2],并输入预训练模型中,以对应模型中的“A、B、A”。
本实施例中,预测病历项目类别指对话文本单元经过预训练模型后的输出,各病历项目类别的预测概率。
403、根据所述标注信息与所述预测病历项目类别,计算所述预训练模型的损失函数;
本实施例中,当标注信息中标注的对话文本单元的病历项目类别与预测病历项目类别相同,则预测正确,否则预测错误。通过预测结果,采用损失函数,计算预训练模型的预测准确性,其中,损失函数可使用交叉熵、KL散度(Kullback–Leibler divergence)、负对数似然(negative log-likelihood)函数等。
优选地,可通过交叉熵作为预训练模型的损失函数,如下所示:
Figure PCTCN2020131681-appb-000001
其中,N为对话文本单元数量,y (i)表示第i个对话文本单元的预测结果(即预测正确或者预测错误,预测正确为1、预测错误为0),
Figure PCTCN2020131681-appb-000002
表示对第i个对话文本单元对各病历项目类别的预测概率,Los表示损失值。
404、对所述预训练模型进行迭代,直到所述损失函数小于预置阈值,得到对话文本分 类模型,其中,所述文本分类模型为病人对话分类模型或医生对话分类模型。
本申请实施例中,详细介绍了病人文本分类模型和医生文本分类模型的生成过程,通过历史对话文本构造历史对话文本单元作为训练样本,以预训练模型训练对对话文本单元的分类,使得电子病历实现自动化生成。
上面对本申请实施例中电子病例自动生成方法进行了描述,下面对本申请实施例中电子病例自动生成装置进行描述,请参阅图5,本申请实施例中电子病例自动生成装置一个实施例包括:
构造模块501,用于获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
分类模块502,用于采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
生成模块503,用于根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
标准化模块504,用于对所述初始电子病例进行标准化处理,得到对应的电子病例。
本申请实施例中,首先,针对在线问诊场景下病人和医生的对话文本,进行病人对话单元、医生对话文本单元构造;然后,将病人对话单元、医生对话文本单元分别输入到对应的病人文本分类模型、医生文本分类模型中进行文本分类;接下来,将通过分类模型得到的分类后的病人的话和医生的话进行病历项目的归档;最后对归类好的病人文本进行标准化,以生成规划范的电子病历。
请参阅图6,本申请实施例中电子病例自动生成装置的另一个实施例包括:
构造模块501,用于获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
分类模块502,用于采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
生成模块503,用于根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
标准化模块504,用于对所述初始电子病例进行标准化处理,得到对应的电子病例。
可选的,在本申请第二方面的第一种实现方式中,所述对话文本包含一个或多个医生对话文本和病人对话文本,所述构造模块501包括:
第一选取单元5011,用于从所述对话文本中选取全部医生对话文本,并将所述各医生对话文本作为医生对话文本单元;
第二选取单元5012,用于依次从所述对话文本中选取一个病人对话文本作为匹配锚点,并选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文;
拼接单元5013,用于将所述医生对话上下文和所述匹配锚点进行拼接,得到对应的病人对话单元。
可选的,在本申请第二方面的第二种实现方式中,所述第二选取单元5012用于:
判断在所述匹配锚点之前是否存在一个或多个医生对话文本以及在所述匹配锚点之后是否存在一个或多个医生对话文本;
若在所述匹配锚点之前和之后均存在一个或多个医生对话文本,则将所述匹配锚点之前和之后的一个或多个医生对话文本作为医生对话上下文;
若在所述匹配锚点之前或之后存在一个或多个医生对话文本,则将所述匹配锚点之前或之后的一个或多个医生对话文本作为医生对话上下文。
可选的,在本申请第二方面的第三种实现方式中,所述标准化模块504包括:
分词单元5041,用于对所述初始电子病历中的对话文本进行分词处理,得到多个分词;
筛除单元5042,用于根据预置词频先验知识,从所述各分词中筛除非医学分词,并将未筛除的分词作为非标准医学分词;
匹配单元5043,用于根据所述非标准医疗分词,遍历预置标准医疗词汇语义库,从所述标准医疗词汇语义库中筛选与所述非标准医疗分词相匹配的标准医疗词汇;
替换单元5044,用于通过所述标准医疗词汇替换对应的非标准医疗分词,得到电子病历。
可选的,在本申请第二方面的第四种实现方式中,所述电子病例自动生成装置还包括模型生成模块505,所述模型生成模块505包括:
构造单元5051,用于获取历史对话文本,并根据所述历史对话文本,构造多个历史对话文本单元,其中,所述历史对话文本单元包括历史病人对话单元和历史医生对话文本单元;
标注单元5052,用于对所述历史对话文本单元的病历项目类别进行标注,得到对应标注信息;
训练单元5053,用于采用预置预训练模型对所述历史对话文本单元进行训练,输出对应的预测病历项目类别;
计算单元5054,用于根据所述标注信息与所述预测病历项目类别,计算所述预训练模型的损失函数;
迭代单元5055,用于对所述预训练模型进行迭代,直到所述损失函数小于预置阈值,得到对话文本分类模型,其中,所述文本分类模型为病人对话分类模型或医生对话分类模型。
可选的,在本申请第二方面的第五种实现方式中,在所述生成模块503之前,还包括:
剔除模块506,用于选取所述病历项目类别中的其他项目类别,并剔除所述其他项目类别中的病人对话单元和医生对话单元。
本申请实施例中,详细介绍了病人对话单元和医生对话单元的生成过程,利用医生角色和病人角色的特性构造对应的对话文本单元,其分类准确性更高;再通过电子病历的标准化处理,将非医学词汇剔除,将非标准医疗词汇以标准医疗词汇进行代替,使得电子病历更规范;接着详细介绍了病人文本分类模型和医生文本分类模型的生成过程,通过历史对话文本构造历史对话文本单元作为训练样本,以预训练模型训练对对话文本单元的分类,使得电子病历实现自动化生成。
上面图5和图6从模块化功能实体的角度对本申请实施例中的电子病例自动生成装置进行详细描述,下面从硬件处理的角度对本申请实施例中电子病例自动生成设备进行详细描述。
图7是本申请实施例提供的一种电子病例自动生成设备的结构示意图,该电子病例自动生成设备700可因配置或性能不同而产生比较大的差异,可以包括一个或一个以上处理器(central processing units,CPU)710(例如,一个或一个以上处理器)和存储器720,一个或一个以上存储应用程序733或数据732的存储介质730(例如一个或一个以上海量存储设备)。其中,存储器720和存储介质730可以是短暂存储或持久存储。存储在存储介质730的程序可以包括一个或一个以上模块(图示没标出),每个模块可以包括对电子病例自动生成设备700中的一系列指令操作。更进一步地,处理器710可以设置为与存储介质730通信,在电子病例自动生成设备700上执行存储介质730中的一系列指令操作。
电子病例自动生成设备700还可以包括一个或一个以上电源740,一个或一个以上有线 或无线网络接口750,一个或一个以上输入输出接口760,和/或,一个或一个以上操作系统731,例如Windows Serve,Mac OS X,Unix,Linux,FreeBSD等等。本领域技术人员可以理解,图7示出的电子病例自动生成设备结构并不构成对电子病例自动生成设备的限定,可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件布置。
本申请还提供一种电子病例自动生成设备,所述电子病例自动生成设备包括存储器和处理器,存储器中存储有计算机可读指令,计算机可读指令被处理器执行时,使得处理器执行上述各实施例中的所述电子病例自动生成方法的步骤。
本申请还提供一种计算机可读存储介质,该计算机可读存储介质可以为非易失性计算机可读存储介质,该计算机可读存储介质也可以为易失性计算机可读存储介质,所述计算机可读存储介质中存储有指令,当所述指令在计算机上运行时,使得计算机执行所述电子病例自动生成方法的步骤。
所属领域的技术人员可以清楚地了解到,为描述的方便和简洁,上述描述的系统,装置和单元的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(read-only memory,ROM)、随机存取存储器(random access memory,RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
以上所述,以上实施例仅用以说明本申请的技术方案,而非对其限制;尽管参照前述实施例对本申请进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围。

Claims (20)

  1. 一种电子病例自动生成方法,其中,所述电子病例自动生成方法包括:
    获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
    采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
    根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
    对所述初始电子病例进行标准化处理,得到对应的电子病例。
  2. 根据权利要求1所述的电子病例自动生成方法,其中,所述对话文本包含一个或多个医生对话文本和病人对话文本,所述根据所述对话文本,构造多个对话文本单元包括:
    从所述对话文本中选取全部医生对话文本,并将所述各医生对话文本作为医生对话文本单元;
    依次从所述对话文本中选取一个病人对话文本作为匹配锚点,并选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文;
    将所述医生对话上下文和所述匹配锚点进行拼接,得到对应的病人对话单元。
  3. 根据权利要求2所述的电子病例自动生成方法,其中,所述选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文包括:
    判断在所述匹配锚点之前是否存在一个或多个医生对话文本以及在所述匹配锚点之后是否存在一个或多个医生对话文本;
    若在所述匹配锚点之前和之后均存在一个或多个医生对话文本,则将所述匹配锚点之前和之后的一个或多个医生对话文本作为医生对话上下文;
    若在所述匹配锚点之前或之后存在一个或多个医生对话文本,则将所述匹配锚点之前或之后的一个或多个医生对话文本作为医生对话上下文。
  4. 根据权利要求1所述的电子病例自动生成方法,其中,所述对所述初始电子病例进行标准化处理,得到对应的电子病例包括:
    对所述初始电子病历中的对话文本进行分词处理,得到多个分词;
    根据预置词频先验知识,从所述各分词中筛除非医学分词,并将未筛除的分词作为非标准医学分词;
    根据所述非标准医疗分词,遍历预置标准医疗词汇语义库,从所述标准医疗词汇语义库中筛选与所述非标准医疗分词相匹配的标准医疗词汇;
    通过所述标准医疗词汇替换对应的非标准医疗分词,得到电子病历。
  5. 根据权利要求1-4中任一项所述的电子病例自动生成方法,其中,所述病人对话分类模型和所述医生对话分类模型的生成过程包括:
    获取历史对话文本,并根据所述历史对话文本,构造多个历史对话文本单元,其中,所述历史对话文本单元包括历史病人对话单元和历史医生对话文本单元;
    对所述历史对话文本单元的病历项目类别进行标注,得到对应标注信息;
    采用预置预训练模型对所述历史对话文本单元进行训练,输出对应的预测病历项目类别;
    根据所述标注信息与所述预测病历项目类别,计算所述预训练模型的损失函数;
    对所述预训练模型进行迭代,直到所述损失函数小于预置阈值,得到对话文本分类模型,其中,所述文本分类模型为病人对话分类模型或医生对话分类模型。
  6. 根据权利要求5所述的电子病例自动生成方法,其中,在所述根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电 子病例之前,还包括:
    选取所述病历项目类别中的其他项目类别,并剔除所述其他项目类别中的病人对话单元和医生对话单元。
  7. 一种电子病例自动生成装置,其中,所述电子病例自动生成装置包括:
    构造模块,用于获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
    分类模块,用于采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
    生成模块,用于根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
    标准化模块,用于对所述初始电子病例进行标准化处理,得到对应的电子病例。
  8. 根据权利要求7所述的电子病例自动生成装置,其中,所述对话文本包含一个或多个医生对话文本和病人对话文本,所述构造模块包括:
    第一选取单元,用于从所述对话文本中选取全部医生对话文本,并将所述各医生对话文本作为医生对话文本单元;
    第二选取单元,用于依次从所述对话文本中选取一个病人对话文本作为匹配锚点,并选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文;
    拼接单元,用于将所述医生对话上下文和所述匹配锚点进行拼接,得到对应的病人对话单元。
  9. 一种电子病例自动生成设备,其中,所述电子病例自动生成设备包括:存储器和至少一个处理器;
    所述至少一个处理器调用所述存储器中的所述指令,以使得所述电子病例自动生成设备执行如下所述的电子病例自动生成方法的步骤:
    获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
    采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
    根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
    对所述初始电子病例进行标准化处理,得到对应的电子病例。
  10. 根据权利要求9所述的电子病例自动生成方法,其中,所述电子病历生成设备被所述处理器执行所述对话文本包含一个或多个医生对话文本和病人对话文本,所述根据所述对话文本,构造多个对话文本单元的步骤时,包括:
    从所述对话文本中选取全部医生对话文本,并将所述各医生对话文本作为医生对话文本单元;
    依次从所述对话文本中选取一个病人对话文本作为匹配锚点,并选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文;
    将所述医生对话上下文和所述匹配锚点进行拼接,得到对应的病人对话单元。
  11. 根据权利要求10所述的电子病例自动生成方法,其中,所述电子病历生成设备被所述处理器执行所述选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文的步骤时,包括:
    判断在所述匹配锚点之前是否存在一个或多个医生对话文本以及在所述匹配锚点之后是否存在一个或多个医生对话文本;
    若在所述匹配锚点之前和之后均存在一个或多个医生对话文本,则将所述匹配锚点之 前和之后的一个或多个医生对话文本作为医生对话上下文;
    若在所述匹配锚点之前或之后存在一个或多个医生对话文本,则将所述匹配锚点之前或之后的一个或多个医生对话文本作为医生对话上下文。
  12. 根据权利要求9所述的电子病例自动生成方法,其中,所述电子病历生成设备被所述处理器执行所述对所述初始电子病例进行标准化处理,得到对应的电子病例的步骤时,包括:
    对所述初始电子病历中的对话文本进行分词处理,得到多个分词;
    根据预置词频先验知识,从所述各分词中筛除非医学分词,并将未筛除的分词作为非标准医学分词;
    根据所述非标准医疗分词,遍历预置标准医疗词汇语义库,从所述标准医疗词汇语义库中筛选与所述非标准医疗分词相匹配的标准医疗词汇;
    通过所述标准医疗词汇替换对应的非标准医疗分词,得到电子病历。
  13. 根据权利要求9-12中任一项所述的电子病例自动生成方法,其中,所述电子病历生成设备被所述处理器执行所述病人对话分类模型和所述医生对话分类模型的生成过程的步骤时,包括:
    获取历史对话文本,并根据所述历史对话文本,构造多个历史对话文本单元,其中,所述历史对话文本单元包括历史病人对话单元和历史医生对话文本单元;
    对所述历史对话文本单元的病历项目类别进行标注,得到对应标注信息;
    采用预置预训练模型对所述历史对话文本单元进行训练,输出对应的预测病历项目类别;
    根据所述标注信息与所述预测病历项目类别,计算所述预训练模型的损失函数;
    对所述预训练模型进行迭代,直到所述损失函数小于预置阈值,得到对话文本分类模型,其中,所述文本分类模型为病人对话分类模型或医生对话分类模型。
  14. 根据权利要求13所述的电子病例自动生成方法,其中,所述电子病历生成设备被所述处理器执行在所述根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例的步骤之前,还包括:
    选取所述病历项目类别中的其他项目类别,并剔除所述其他项目类别中的病人对话单元和医生对话单元。
  15. 一种计算机可读存储介质,所述计算机可读存储介质上存储有指令,其中,所述指令被处理器执行时实现如下所述的电子病例自动生成方法的步骤:
    获取线上问诊场景下的对话文本,并根据所述对话文本,构造多个对话文本单元,其中,所述对话文本单元包括病人对话单元和医生对话文本单元;
    采用预置病人对话分类模型、预置医生对话分类模型分别对所述各病人对话单元和所述各医生对话单元进行分类,得到对应的病历项目类别;
    根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例;
    对所述初始电子病例进行标准化处理,得到对应的电子病例。
  16. 根据权利要求15所述的电子病例自动生成方法,其中,所述电子病例自动生成的指令被所述处理器执行所述对话文本包含一个或多个医生对话文本和病人对话文本,所述根据所述对话文本,构造多个对话文本单元的步骤时,包括:
    从所述对话文本中选取全部医生对话文本,并将所述各医生对话文本作为医生对话文本单元;
    依次从所述对话文本中选取一个病人对话文本作为匹配锚点,并选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文;
    将所述医生对话上下文和所述匹配锚点进行拼接,得到对应的病人对话单元。
  17. 根据权利要求16所述的电子病例自动生成方法,其中,所述电子病例自动生成的指令被所述处理器执行所述选取所述匹配锚点之前和/或之后的,一个或多个医生对话文本并作为医生对话上下文的步骤时,包括:
    判断在所述匹配锚点之前是否存在一个或多个医生对话文本以及在所述匹配锚点之后是否存在一个或多个医生对话文本;
    若在所述匹配锚点之前和之后均存在一个或多个医生对话文本,则将所述匹配锚点之前和之后的一个或多个医生对话文本作为医生对话上下文;
    若在所述匹配锚点之前或之后存在一个或多个医生对话文本,则将所述匹配锚点之前或之后的一个或多个医生对话文本作为医生对话上下文。
  18. 根据权利要求15所述的电子病例自动生成方法,其中,所述电子病例自动生成的指令被所述处理器执行所述对所述初始电子病例进行标准化处理,得到对应的电子病例的步骤时,包括:
    对所述初始电子病历中的对话文本进行分词处理,得到多个分词;
    根据预置词频先验知识,从所述各分词中筛除非医学分词,并将未筛除的分词作为非标准医学分词;
    根据所述非标准医疗分词,遍历预置标准医疗词汇语义库,从所述标准医疗词汇语义库中筛选与所述非标准医疗分词相匹配的标准医疗词汇;
    通过所述标准医疗词汇替换对应的非标准医疗分词,得到电子病历。
  19. 根据权利要求15-18中任一项所述的电子病例自动生成方法,其中,所述电子病例自动生成的指令被所述处理器执行所述病人对话分类模型和所述医生对话分类模型的生成过程的步骤时,包括:
    获取历史对话文本,并根据所述历史对话文本,构造多个历史对话文本单元,其中,所述历史对话文本单元包括历史病人对话单元和历史医生对话文本单元;
    对所述历史对话文本单元的病历项目类别进行标注,得到对应标注信息;
    采用预置预训练模型对所述历史对话文本单元进行训练,输出对应的预测病历项目类别;
    根据所述标注信息与所述预测病历项目类别,计算所述预训练模型的损失函数;
    对所述预训练模型进行迭代,直到所述损失函数小于预置阈值,得到对话文本分类模型,其中,所述文本分类模型为病人对话分类模型或医生对话分类模型。
  20. 根据权利要求20所述的电子病例自动生成方法,其中,所述电子病例自动生成的指令被所述处理器执行在所述根据所述病历项目类别,依次将所述各病人对话单元和所述各医生对话单元进行归档操作,生成对应的初始电子病例的步骤之前,还包括:
    选取所述病历项目类别中的其他项目类别,并剔除所述其他项目类别中的病人对话单元和医生对话单元。
PCT/CN2020/131681 2020-09-07 2020-11-26 电子病例自动生成方法、装置、设备及存储介质 WO2021208444A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010926413.8 2020-09-07
CN202010926413.8A CN112017744A (zh) 2020-09-07 2020-09-07 电子病例自动生成方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2021208444A1 true WO2021208444A1 (zh) 2021-10-21

Family

ID=73515979

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/131681 WO2021208444A1 (zh) 2020-09-07 2020-11-26 电子病例自动生成方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN112017744A (zh)
WO (1) WO2021208444A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114356860A (zh) * 2022-01-06 2022-04-15 支付宝(杭州)信息技术有限公司 对话生成方法及装置
CN115208846A (zh) * 2022-07-08 2022-10-18 武汉联影医疗科技有限公司 一种会话交互方法和系统

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112036154B (zh) * 2020-08-31 2023-05-23 康键信息技术(深圳)有限公司 基于问诊对话的电子病历生成方法、装置和计算机设备
CN113327657B (zh) * 2021-05-27 2023-08-25 挂号网(杭州)科技有限公司 病例报告生成方法、装置、电子设备以及存储介质
CN117912625B (zh) * 2024-03-20 2024-05-28 广州源康健信息科技有限公司 一种基于问诊对话的电子病历生成方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106095913A (zh) * 2016-06-08 2016-11-09 广州同构医疗科技有限公司 一种电子病历文本结构化方法
CN106951703A (zh) * 2017-03-15 2017-07-14 长沙富格伦信息科技有限公司 一种生成电子病历的系统及方法
CN107038336A (zh) * 2017-03-21 2017-08-11 科大讯飞股份有限公司 一种电子病历自动生成方法及装置
CN108182262A (zh) * 2018-01-04 2018-06-19 华侨大学 基于深度学习和知识图谱的智能问答系统构建方法和系统
US20190347269A1 (en) * 2018-05-08 2019-11-14 Siemens Healthcare Gmbh Structured report data from a medical text report

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104485105B (zh) * 2014-12-31 2018-04-13 中国科学院深圳先进技术研究院 一种电子病历生成方法和电子病历系统
CN108899064A (zh) * 2018-05-31 2018-11-27 平安医疗科技有限公司 电子病历生成方法、装置、计算机设备和存储介质
KR102111775B1 (ko) * 2018-06-25 2020-05-15 서울대학교 산학협력단 진료 데이터 수집 관리 시스템 및 방법
CN109859811A (zh) * 2018-12-29 2019-06-07 北京天鹏恒宇科技发展有限公司 一种门诊病历智能输入系统
CN109887557A (zh) * 2018-12-29 2019-06-14 北京天鹏恒宇科技发展有限公司 一种智能语音预问诊系统
JP2020113004A (ja) * 2019-01-10 2020-07-27 エヌ・ティ・ティ・コミュニケーションズ株式会社 情報処理装置、電子カルテ作成方法および電子カルテ作成プログラム
CN110610751A (zh) * 2019-09-09 2019-12-24 北京左医科技有限公司 一种电子病历录入系统及方法
CN111180025A (zh) * 2019-12-18 2020-05-19 东北大学 表示病历文本向量的方法、装置及问诊系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106095913A (zh) * 2016-06-08 2016-11-09 广州同构医疗科技有限公司 一种电子病历文本结构化方法
CN106951703A (zh) * 2017-03-15 2017-07-14 长沙富格伦信息科技有限公司 一种生成电子病历的系统及方法
CN107038336A (zh) * 2017-03-21 2017-08-11 科大讯飞股份有限公司 一种电子病历自动生成方法及装置
CN108182262A (zh) * 2018-01-04 2018-06-19 华侨大学 基于深度学习和知识图谱的智能问答系统构建方法和系统
US20190347269A1 (en) * 2018-05-08 2019-11-14 Siemens Healthcare Gmbh Structured report data from a medical text report

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114356860A (zh) * 2022-01-06 2022-04-15 支付宝(杭州)信息技术有限公司 对话生成方法及装置
CN115208846A (zh) * 2022-07-08 2022-10-18 武汉联影医疗科技有限公司 一种会话交互方法和系统
CN115208846B (zh) * 2022-07-08 2023-06-09 武汉联影医疗科技有限公司 一种会话交互方法和系统

Also Published As

Publication number Publication date
CN112017744A (zh) 2020-12-01

Similar Documents

Publication Publication Date Title
WO2021208444A1 (zh) 电子病例自动生成方法、装置、设备及存储介质
CN107731269B (zh) 基于原始诊断数据和病历文件数据的疾病编码方法及系统
CN107705839B (zh) 疾病自动编码方法及系统
US11983210B2 (en) Methods and systems for generating summaries given documents with questions and answers
WO2022095368A1 (zh) 基于文本生成模型的问答语料生成方法和装置
US9779211B2 (en) Computer-assisted abstraction for reporting of quality measures
US20220172725A1 (en) Systems and methods for extracting information from a dialogue
CN110390003A (zh) 基于医疗的问答处理方法及系统、计算机设备及可读介质
CN109857846B (zh) 用户问句与知识点的匹配方法和装置
CN109299227B (zh) 基于语音识别的信息查询方法和装置
US20220366127A1 (en) Legal Document Generation
WO2022068160A1 (zh) 基于人工智能的重症问诊数据识别方法、装置、设备及介质
US11532387B2 (en) Identifying information in plain text narratives EMRs
Selvaraj et al. Medication regimen extraction from medical conversations
CN112655047A (zh) 对医学记录分类的方法
CN117033568A (zh) 一种医学数据指标解读方法、装置、存储介质及设备
CN114360678A (zh) 信息处理方法、装置、设备和存储介质
CN117877660A (zh) 基于语音识别的医学报告获取方法及系统
US20230298589A1 (en) Ai platform for processing speech and video information collected during a medical procedure
CN116719840A (zh) 一种基于病历后结构化处理的医疗信息推送方法
CN115358817A (zh) 基于社交数据的智能产品推荐方法、装置、设备及介质
US20220382982A1 (en) System and method of automatic topic detection in text
CN113761899A (zh) 一种医疗文本生成方法、装置、设备及存储介质
Freitas et al. Machine Learning Powered Data Platform for High-Quality Speech and NLP Workflows.
CN117809800A (zh) 健康指导信息生成方法、装置、电子设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20931269

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20931269

Country of ref document: EP

Kind code of ref document: A1