WO2018205609A1 - 医疗智能分诊方法和设备 - Google Patents

医疗智能分诊方法和设备 Download PDF

Info

Publication number
WO2018205609A1
WO2018205609A1 PCT/CN2017/116370 CN2017116370W WO2018205609A1 WO 2018205609 A1 WO2018205609 A1 WO 2018205609A1 CN 2017116370 W CN2017116370 W CN 2017116370W WO 2018205609 A1 WO2018205609 A1 WO 2018205609A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
disease
patient
medical
triage
Prior art date
Application number
PCT/CN2017/116370
Other languages
English (en)
French (fr)
Inventor
张振中
Original Assignee
京东方科技集团股份有限公司
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 京东方科技集团股份有限公司 filed Critical 京东方科技集团股份有限公司
Priority to EP17908895.0A priority Critical patent/EP3623970A4/en
Priority to US16/069,139 priority patent/US20210280301A1/en
Publication of WO2018205609A1 publication Critical patent/WO2018205609A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
    • G16H40/20—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

  • the present disclosure relates to the field of computer technologies, and in particular, to a medical intelligent triage method, a medical intelligent triage device, and a computer readable storage medium.
  • the doctor's source is limited, and patients who register again may only be able to see the doctor in another day, delaying the golden diagnosis period of the disease.
  • the expert number of the rarer source there is also the problem that the patient is not diagnosed, or only the check-up slip can be issued after many days. In addition to delaying the diagnosis of the disease, these conditions are often accompanied by high time costs and transportation costs.
  • the present disclosure provides a medical intelligent triage method, a medical intelligent triage device, and a computer readable storage medium, which can be based on knowledge mining and semantic relations. (Semantic Relation), through the information provided by the patient to mine relevant medical knowledge from the medical literature, automatically select the interaction with the patient, so as to determine the patient's triage more quickly and accurately.
  • Semantic Relation Through the information provided by the patient to mine relevant medical knowledge from the medical literature, automatically select the interaction with the patient, so as to determine the patient's triage more quickly and accurately.
  • a medical intelligence triage method may include: extracting disease-related symptoms and signs from the patient information as candidate factor information from the patient information by the candidate factor analyzer; and identifying, by the knowledge miner, obtaining a plurality of symptoms related from the medical literature according to the candidate factor information
  • Candidate diseases and treatment measures are used as identification knowledge information; the identification knowledge information is matched with the candidate factor information by a matcher; the above steps are repeated until it is determined that the department to which the disease belongs or the patient information has been extracted and matched, returning to the disease The department is the result of the triage.
  • the patient information may be obtained by a patient's verbal expression language or electronic entry through a human-computer interaction interface, and returning the determined department as a triage result may include determining by a human-computer interaction interface voice broadcast determined by a department or display. Electronic information text of the department.
  • the candidate factor information may include keywords or key phrases and time events of symptoms and signs associated with the disease
  • the step of extracting the candidate factor information may include: passing the natural factor by the candidate factor analyzer
  • the language processing technique extracts the keywords or key phrases and time events from the textual information provided by the patient from a spoken word provided by the patient or by an information extraction technique.
  • the step of acquiring the identification knowledge information may include: the content retriever retrieving the document content related to the keyword or the key phrase from the medical literature according to the keyword or the key phrase in the candidate factor information;
  • the knowledge extractor uses natural language processing techniques to find and mine disease-related knowledge from retrieved document content.
  • the step of retrieving document content related to a keyword or a key phrase from the medical literature may include the content retriever using a natural language processing technique to perform segmentation and named entity recognition of the medical document, and establishing an inverted row An index; and the content retriever retrieves a document containing all keywords or key phrases from the inverted index according to keywords or key phrases in the candidate factor information.
  • the natural language processing technique is used to search from the retrieved document content.
  • the step of finding and mining knowledge related to the disease may include: the knowledge extractor determining a segment of the keyword or key phrase in the document, and extracting candidate diseases related to symptoms and signs by semantic relationship.
  • the step of matching the identification knowledge information with the candidate factor information may include: if there is only one candidate disease output, determining the disease and the corresponding belonging department according to the candidate factor; If there are multiple candidate diseases to be output, the belongings are determined according to the prevalence of the disease with the same symptoms and signs in the set time period or the set region; or, further, different diseases with cross-symptoms and signs are selected.
  • the most discriminating information is used as an extension problem, further obtaining patient information and extracting new candidate factor information for further matching process.
  • the information gain can be used to determine the degree of discrimination between different diseases with cross-symptoms and signs. The greater the information gain, the greater the discrimination, and the smaller the information gain, the smaller the discrimination, and the information gain calculation.
  • the formula can be:
  • Symptom means symptoms
  • Disease means disease
  • H(.) means entropy
  • a medical intelligence triage device including a candidate factor analyzer, an authentication knowledge miner, and a matcher.
  • the candidate factor analyzer can be configured to extract disease-related symptoms and signs from the patient information as candidate factor information.
  • the authentication knowledge miner is coupled to the candidate factor analyzer, and is configured to obtain, from the medical literature, a plurality of candidate diseases related to symptoms and treatment measures as the identification knowledge information according to the candidate factor information.
  • the matcher is connected to the authentication knowledge miner, and can be configured to match the authentication knowledge information with the candidate factor information to determine a department to which the disease belongs, and return to the department to which the disease belongs as a triage result.
  • the device may further include a human-machine interaction facility, and the human-machine interaction facility may be separately connected to the candidate factor analyzer and the matcher.
  • the human-machine interaction facility provides a human-computer interaction interface for the patient and can be configured to collect patient information and present the triage results to the patient.
  • the patient information may be obtained by a patient's verbal expression language or electronic entry, and the triage results may be returned to the patient via a voice announcement or an electronic message display.
  • the candidate factor information may include keywords or key phrases and time events of symptoms and signs associated with the disease
  • the candidate factor analyzers may be configured to provide oral presentations from the patient through natural language processing techniques
  • the keyword or key phrase and time event are extracted from the information text provided by the patient in the expression language or by information extraction techniques.
  • the authentication knowledge miner can include a content retriever and a knowledge extractor.
  • the content retriever may be configured to retrieve document content related to keywords or key phrases from the medical literature based on keywords or key phrases output by the candidate factor analyzer.
  • the knowledge extractor can be configured to utilize natural language processing techniques to find and mine disease-related knowledge from retrieved document content.
  • the content retriever can be configured to: perform segmentation and named entity recognition of the medical document using natural language processing techniques, and establish an inverted index; and keywords or key outputs according to the candidate factor analyzer A phrase that retrieves documents containing all keywords or key phrases from the inverted index.
  • the knowledge extractor can be configured to determine a segment of a keyword or key phrase in the document and to extract candidate diseases associated with symptoms and signs by semantic relationships.
  • the matcher can be configured to determine the disease and the corresponding affiliated department based on the candidate factors. If there are multiple candidate diseases to be output, the matcher may be configured to determine the belonging department according to the popularity of the disease having the same symptoms and signs over a set period of time or a set region; or, the match The device can be configured to further select the most discriminating information between the different diseases with crossed symptoms and signs as an extension problem, further obtaining patient information and extracting new candidate factor information for further matching.
  • the matcher can be configured to determine the degree of discrimination between different diseases having cross-symptoms and signs by information gain. The greater the information gain, the greater the discrimination and the smaller the information gain. The smaller the degree of discrimination.
  • the information gain calculation formula can be:
  • Symptom means symptoms
  • Disease means disease
  • H(.) means entropy
  • a medical intelligent triage device comprising: One or more processors and memories.
  • the memory stores computer executable instructions configured to perform any of the methods described above when executed by the one or more processors.
  • a computer readable storage medium having computer executable instructions thereon, when executed by one or more processors, causing the one or more processors Perform any of the methods described above.
  • Some embodiments of the present disclosure may achieve at least one of the following beneficial effects and/or other benefits: the medical intelligence triage method and the medical intelligence triage device by analyzing the interaction content of the patient, from medical literature or other medicine Relevant knowledge and rules are found in the data to form medical evidence for identifying knowledge information, and then to analyze the patient's condition and give relevant recommendations. Therefore, according to the main symptoms and signs of the patient, the possible diseases and their affiliated specialties are matched, and the purpose of the effective visiting department or medical treatment path is recommended.
  • FIG. 1 is a flow chart of a medical smart triage method in accordance with one embodiment
  • FIG. 2 is a schematic diagram of the architecture of a medical intelligence triage device according to an embodiment
  • FIG. 3 is a schematic diagram of an example of a triage according to an embodiment
  • the medical intelligent triage refers to determining a possible disease according to the main symptoms and signs of the patient, determining the priority of the condition and its affiliated specialty, and recommending an effective treatment path.
  • the medical intelligence triage method according to some embodiments and the corresponding medical intelligence triage device are based on knowledge mining and Semantic Relation, and the relevant medical knowledge is automatically extracted from the medical literature by the information provided by the patient, automatically Select the interaction with the patient's disease to determine the patient's triage more quickly and accurately.
  • the current method of triage is often accompanied by high time cost and transportation cost.
  • the above-mentioned medical intelligent triage method realizes the automatic selection and interaction of patients through triage based on knowledge mining. To determine the patient's triage department more quickly and accurately.
  • FIG. 1 shows a flow chart of a medical intelligence triage method in accordance with one embodiment.
  • the medical intelligence triage method may include steps S1-S5.
  • Step S1) Collect patient information through a human-computer interaction interface.
  • the patient information can be obtained by the patient's verbal expression language or electronic entry through the human-computer interaction interface, and the triage result can also be returned to the patient through the voice broadcast or electronic information text display using the human-computer interaction interface.
  • patient information can be collected in a variety of ways, such as human-computer interaction or voice interaction, to obtain as much information as possible of the patient including the main symptoms.
  • the patient information can be acquired in one time for subsequent analysis and matching processing; it can also be obtained step by step, and gradually provide more new patient information for analysis and matching processing.
  • Step S2) Extracting disease-related symptoms and signs from the patient information as candidate factor information by the candidate factor analyzer.
  • the candidate factor analyzer analyzes the information provided by the patient and extracts from it.
  • the main symptoms and major signs associated with the disease are used as candidate information.
  • keywords or key phrases and time events of symptoms and signs associated with the disease may be extracted from the textual information provided by the patient from a patient-provided verbal representation language by natural language processing techniques or by information extraction techniques. The above keywords or key phrases and time events can be used as candidate factor information.
  • Step S3) The identification knowledge miner obtains a plurality of candidate diseases related to the symptoms and treatment measures from the medical literature as the identification knowledge information according to the candidate factor information.
  • the identification knowledge miner can extract knowledge about the diseases associated with the main symptoms and the treatment measures from the mass medical literature. That is, the document content related to the keyword or the key phrase can be retrieved from the medical literature according to the keyword or the key phrase in the candidate factor information; and the natural language processing technology is used to find and mine the disease-related knowledge from the retrieved document content. .
  • the step of obtaining the knowledge information related to the main symptom related diseases and the treatment measures from the medical literature may further include:
  • Step S4) matching the identification knowledge information with the candidate factor information by the matcher.
  • the matcher extracts medical evidence from the knowledge information related to the main symptom-related diseases and treatment measures acquired in step S3), and matches the candidate factor information according to the medical evidence.
  • the match may include the following:
  • the candidate disease and the corresponding affiliated department are determined according to the candidate factors
  • the method of obtaining patient information can further ask the patient for consultation, and the purpose is to further obtain information related to the disease, and determine the result of the triage according to the candidate factors related to the disease and the information of the possible diseases, and obtain the result of the triage.
  • the information gain can be used to determine the degree of discrimination between different diseases with cross-symptoms and signs. The greater the information gain, the greater the discrimination, and the smaller the information gain, the smaller the discrimination, and the information gain calculation.
  • the formula can be:
  • Symptom means symptoms
  • Disease means disease
  • H(.) means entropy
  • the actual most likely triage results can be obtained by further screening and locating multiple candidate diseases.
  • Step S5) repeat the above steps until it is determined that the department to which the disease belongs or the patient information has been extracted and matched, and returned to the department belonging to the disease as a result of the triage.
  • Steps S1) - S4) are repeated until the department to which the disease belongs is determined, or any disease-related information can no longer be obtained from the patient information, returning to the department to which the disease belongs and showing the patient the results of the triage that may be ill.
  • a medical intelligence triage device is also provided. Also based on knowledge mining for triage, the triage device is able to automatically select candidate disease information related to the disease in the patient information to determine the patient's triage department more quickly and accurately.
  • the medical intelligent triage device may include a human-machine interaction facility 1, a candidate factor analyzer 2, an authentication knowledge miner 3, and a matcher 4.
  • the human-computer interaction facility 1 can provide a human-computer interaction interface for the patient 5, can be configured to collect patient information and present the triage results to the patient 5.
  • the patient information can be obtained by the patient's verbal expression language or electronic entry, and the triage result can be returned to the patient by voice broadcast or electronic information text display.
  • the candidate factor analyzer 2 is coupled to the human interaction facility 1 and can be configured to extract disease-related symptoms and signs from the patient information as candidate factor information.
  • the candidate factor analyzer 2 may be configured to analyze information provided by the patient 5 collected by the human-machine interaction facility 1 from which the main symptoms and major signs associated with the disease are extracted as candidate factor information.
  • the candidate factor analyzer 2 may be configured to extract symptoms and signs associated with the disease from a patient-provided text in the oral presentation language provided by the patient or through an information extraction technique by natural language processing techniques. Key words or key phrases and time events, the above keywords or key phrases and time events can be used as candidate factor information.
  • Disease candidate factors can be obtained by the candidate factor analyzer 2 extracting disease-related information in the patient information.
  • the authentication knowledge miner 3 is connected to the candidate factor analyzer 2, and may be configured to acquire a plurality of candidate diseases related to symptoms and treatment measures from the medical literature as the identification knowledge information based on the candidate factor information. Based on the information extracted from the candidate factor analyzer 2, the knowledge related to the main symptom-related diseases and the treatment measures can be extracted from the mass medical literature 6, and information on possible diseases corresponding to the disease candidate factors can be obtained.
  • the authentication knowledge miner 3 can include a content retriever and a knowledge extractor.
  • the content retriever can be configured to retrieve document content related to keywords or key phrases from the medical literature based on keywords or key phrases output by the candidate factor analyzer.
  • the knowledge extractor can be configured to utilize natural language processing techniques to find and mine disease-related knowledge from retrieved document content.
  • the content retriever can be configured to perform segmentation and named entity recognition of the medical document using natural language processing techniques, and establish an inverted index; and retrieve the inverted index from the keywords or key phrases output by the candidate factor analyzer A document containing all keywords or key phrases.
  • the knowledge extractor can be configured to determine the segment of the keyword or key phrase in the document and to extract candidate diseases associated with symptoms and signs through semantic relationships.
  • the matcher 4 can be connected to the human-machine interaction facility 1 and the authentication knowledge miner 3, respectively, and can be configured to match the authentication knowledge information with the candidate factor information and return the triage result.
  • the triage result of the disease that the patient 5 may suffer can be matched by the comparison of the knowledge and candidate factor information output from the identification knowledge miner 3. If there is only one candidate disease output, the matcher 4 can be configured to determine the candidate disease and the corresponding affiliated department based on the candidate factors.
  • the matcher 4 can be configured It is used to determine the department according to the prevalence of the disease with the same symptoms and signs in the set time period or the set region; or, to further select the information with the highest degree of discrimination between different diseases with cross-symptoms and signs as Extend the problem, further obtain patient information and extract new candidate information for further matching.
  • the matcher 4 can judge the degree of discrimination between different diseases having cross-symptoms and signs by the information gain. The larger the information gain, the larger the discrimination, and the smaller the information gain, the smaller the discrimination.
  • the information gain calculation formula can be:
  • Symptom means symptoms
  • Disease means disease
  • H(.) means entropy
  • the medical intelligence triage device can also be implemented by one or more processors and memories.
  • the memory stores computer executable instructions configured to perform any of the methods described above when executed by the one or more processors.
  • a computer readable storage medium having computer executable instructions thereon, when executed by one or more processors, causing the one or more processors Perform any of the methods described above.
  • FIG. 3 shows a schematic diagram of a triage example in accordance with one embodiment.
  • the function and implementation manner of the above medical intelligent triage method and medical intelligent triage device modules will be described in detail below with reference to FIG. 3 in combination with a triage example.
  • the correspondence between the four modules in the medical intelligent triage device and the five major steps in the medical intelligent triage method is that the human-computer interaction facility 1 involves step S1) and step S5), the candidate factor analyzer 2 Involving step S2), the authentication knowledge miner 3 involves step S3) and the matcher 4 involves step S4).
  • the human-computer interaction interface 1 is provided with a human-computer interaction interface for the patient 5, which can be configured to collect information such as symptoms and signs provided by the patient 5.
  • the patient information can be formed by asking the patient 5 for an uncomfortable site or providing corresponding symptoms, signs, and the like, and then collecting the reply information of the patient 5.
  • the human-machine interaction facility 1 can also be configured to present to the patient 5 the final triage results of the medical intelligence triage device.
  • the candidate information analyzer 1 analyzes the patient information collected by the human-machine interaction facility 1 from which main information such as symptoms and signs related to the disease is extracted.
  • keywords or key phrases key words or key phrases herein refer to disease symptoms and signs, etc.
  • time events may be extracted from patient information speech or text by natural language processing techniques and information extraction techniques.
  • the device asks "Where are you uncomfortable?"
  • Patient 5 can answer "He started to have a stomachache yesterday, and today the waist is beginning to hurt.”
  • the Candidate Factor Analyzer 2 can automatically extract the keyword “stomach pain” and the key phrase “waist also begins to hurt”.
  • the time nouns "yesterday” and “today” can be extracted, and then the time, the event (yesterday, stomach pain) and (today, stomach pain and low back pain) can be analyzed.
  • the relevant knowledge information is extracted from the mass medical document 6 by the knowledge crawler 3 based on the information extracted by the candidate factor analyzer 2.
  • the authentication knowledge miner 3 can include a content retriever and a knowledge extractor.
  • the content retriever can retrieve related content from the medical document 6 based on keywords or key phrases output by the candidate factor analyzer 2.
  • the following process may be included: first, the medical document 6 is subjected to word segmentation and named entity recognition using a natural language processing technique, and an inverted index is established, and the inverted index may take the form: "word 1" - "document 1, document i, ...., document N", wherein "document 1, document i, document N" are documents containing "word 1"; and then according to keywords or key phrases output by the candidate factor analyzer 2, retrieve documents containing all keywords or key phrases from the inverted index.
  • w1, w2, ..., wK is K keywords or key phrases
  • S1, S2, ..., SK is the corresponding document collection in the inverted index (ie, S1 is a document collection containing w1, And so on)
  • the knowledge extractor can exploit the natural language processing techniques to mine relevant knowledge from the retrieved documents.
  • a process may be included that determines a segment of a keyword or key phrase in the document, and a candidate disease associated with symptom signs (keywords or key phrases such as headaches, dizziness, etc.) by semantic relationships. For example, given the keywords "headache” and "dizziness", it is determined from the retrieved document that "high blood pressure can cause arterial congestion, dilatation, headache, and sometimes nausea, vomiting, dizziness.”
  • the knowledge extractor can obtain the following semantic relationship: high blood pressure causes headache, high blood Pressure causes nausea, high blood pressure causes vomiting, high blood pressure causes dizziness, and so on.
  • Hypertension is a candidate disease that causes the patient to have 5 headaches and vertigo.
  • the triage of the possible diseases of the patient 5 is analyzed by the matcher 4 from the knowledge and candidate factor information outputted by the identification knowledge miner 3.
  • the analysis can include the following process:
  • the candidate knowledge miner 3 If the candidate knowledge miner 3 outputs only one candidate disease, the disease and the corresponding department are output. For example, if the candidate disease outputted by the knowledge miner 3 is only "hypertension” for "headache” and “dizziness”, “hypertension” and the department "cardiovascular department” to be registered are output;
  • the information with the highest degree of discrimination can be selected as the question to consult the patient 5.
  • the degree of discrimination is determined by the information gain. The larger the information gain is, the larger the discrimination is, and the smaller the information gain is, the smaller the discrimination is.
  • Table 1 the returned knowledge is shown in Table 1:
  • Trigger Headache dizziness, nausea, urine Migraine Trigger Headache, dizziness, nausea, vomiting Neurasthenia Trigger Headache, dizziness, insomnia, anxiety, irritability
  • the device needs to communicate with the patient 5 through the human-machine interaction device 1 to collect new patient information to determine the symptoms or signs of the disease that may be suffered.
  • the device can ask the question "Do you have any symptoms of nausea?" or "Do you have recently lost sleep?” or any other related symptoms. Since the types and symptoms are usually many, it is crucial that the device selects symptoms in order to determine the patient 5's disease information as quickly as possible.
  • the medical intelligent triage method of the present embodiment regards the prevalence of the disease as a prior probability, and in the case where the other conditions are the same, the symptoms of the disease with a high prevalence are preferentially selected. For example, under the above-mentioned disease prevalence ratio conditions, the possible diseases in Table 1 are high blood pressure;
  • a patient with symptoms of “headache and dizziness” is assumed to be diagnosed.
  • the device cannot determine whether the patient is classified as “hypertension” or “migraine” or “distressed”. Because these three diseases can cause the above symptoms.
  • the device needs to collect more patient information.
  • the device first judges which disease has the highest number of patients according to the prevalence of the disease, for example, in this example, the proportion of patients who assume hypertension is 70% (this percentage can be obtained by statistically diagnosed medical records of the hospital), ie In most cases, patients with symptoms of "headache, dizziness" are suffering from high blood pressure. At this time, the device selects a symptom from the hypertension-related symptoms obtained by the identification knowledge miner 3 to inquire the patient to determine whether the patient has high blood pressure.
  • IG information gain
  • the information gain formula based on the degree of discrimination of the symptoms is:
  • the device chooses the symptom "more urine” to get further information or questions from the patient (for example, have you had a lot of urine recently?). If the information or answer provided by the patient is "Yes”, the result of the triage diagnosis is determined. If the answer is "No”, the results of the hypertension triage are excluded and the remaining diseases are considered as candidate diseases. It can be seen that for this example, the triage matching is to repeat the above steps for "migraine” and "neurasthenia” until the disease triage is determined or the patient terminates the procedure.
  • the medical intelligent triage method of the present embodiment and the corresponding medical intelligent triage device first select the most popular disease by the prevalence of the disease, and then select the most differentiated symptom from the disease.
  • the patient 5 is inquired through the human-machine interaction device 1. For example, for the above example, first choose hypertension as the most likely patient 5 Disease (because the prevalence of hypertension is up to 70%), and then from the knowledge output from the knowledge extractor, look for other symptoms of high blood pressure, choose the most differentiated symptom "more urine", and cause problems "Do you have more urine recently?
  • the inquiry is made to the patient 5 through the human-machine interaction device 1.
  • the cycle is repeated until the disease triage is determined or new disease candidate information can no longer be obtained.
  • the device-matched triage result is presented to the user through the human-machine interaction device 1.
  • the medical intelligent triage method and the medical intelligent triage device find relevant medical knowledge and laws from medical literature or other medical data through analysis and patient interaction content to form medical evidence for identifying knowledge information, and then analyze the patient's disease condition and Give relevant advice. Therefore, according to the main symptoms and signs of the patient, the possible diseases and their affiliated specialties are matched, and the purpose of the effective visiting department or medical treatment path is recommended.
  • the foregoing embodiment is only exemplified by the division of the foregoing functional modules.
  • the foregoing functions may be allocated to different functional modules as needed.
  • the internal structure of the device can be divided into different functional modules to perform all or part of the functions described above.
  • the function of one module described above may be completed by multiple modules, and the functions of the above multiple modules may also be integrated into one module.
  • any reference signs placed in parentheses shall not be construed as limiting the claim.
  • the word “comprising” does not exclude the presence of the elements or the The word “a” or “an” or “an”
  • the invention can be hard by means of several separate elements
  • the implementation can also be implemented by appropriately programmed software or firmware, or by any combination thereof.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Biomedical Technology (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • General Business, Economics & Management (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Artificial Intelligence (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Algebra (AREA)
  • Software Systems (AREA)
  • Operations Research (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种医疗智能分诊方法和设备以及计算机可读存储介质，涉及计算机技术领域。该医疗智能分诊方法，包括：从患者信息中抽取与疾病相关的症状和体征作为候选因素信息；根据候选因素信息，从医学文献中获取与症状相关的多种候选疾病以及治疗措施作为鉴定知识信息；将鉴定知识信息与候选因素信息进行匹配；重复上述步骤，直到确定疾病所属科室或者患者信息已抽取并匹配完毕，返回疾病所属科室作为分诊结果。该方法和设备通过分析和患者的交互内容，从医学文献或者其他医学数据中发现相关知识和规律从而形成鉴定知识信息的医学证据，进而分析患者的患病情况并给出相关建议。从而达到根据患者的主要症状及体征，匹配出可能的疾病及其隶属专科，并推荐有效的就诊的科室或就医路径的目的。

Description

医疗智能分诊方法和设备
相关申请
本申请要求2017年5月12日提交、申请号为201710333821.0的中国专利申请的优先权，该申请的全部内容通过引用并入本文。
技术领域
本公开涉及计算机技术领域，具体涉及一种医疗智能分诊方法、医疗智能分诊设备以及计算机可读存储介质。
背景技术
在医院门诊进行分诊时，准确、快速地确定患者所需要的科室是患者所热切期望的。然而现实中的门诊分诊场景是，患者对医院的科室分类不甚了解，只能通过在挂号时简单咨询挂号窗口或者分诊台的工作人员挂什么科比较合适。然而术业有专攻，很多挂号窗口或分诊台工作人员并非医师专业出身。因此，患者面临的结果往往是，好不容易挂到的号，见到医生却被告知病症和所挂科室并不符合，要对症看病需要重新选择另一个科室挂号。这时，患者要么重新挂号，要么请医生开大概对症的药了事。然而医生的号源有限，再次挂号患者可能只能改天才能看病，耽误病情的黄金诊断期。对于更稀少号源的专家号，同样存在病不对诊，或者仅能开具多日后的检查预约单的问题。这些情况除了耽误病情诊断，往往还伴随着很高的时间成本和交通成本。
上述情况距离科学分诊还有很大的距离。目前的分诊设备通常基于人工编写的规则库，需要耗费大量的人力和时间，适用的范围窄。随着人工智能的兴起和人文关怀的普及，医疗智能分诊(Intelligent Triage)开始发展起来。相比传统的门诊分诊，医疗智能分诊能够更快速、更准确的判断疾病并给出合理建议。例如，一种提出的分诊方法是基于深度学习方法的Airdoc的分诊方式，使用深度学习方法通常是端到端(end to end)的应用，其过程对于人来说具有不可解释性，并且缺少良好的信息交互，这对于医学应用来说是个很大的弊端。
发明内容
为了解决或缓解上述现有技术中的至少一个缺陷，本公开提供了一种医疗智能分诊方法、医疗智能分诊设备以及计算机可读存储介质，它们能基于知识挖掘(Knowledge Mining)和语义关系(Semantic Relation)，通过患者提供的信息从医学文献中挖掘相关的医学知识，自动地选择和患者的交互内容，从而更快更准地确定患者的分诊情况。
根据本公开的一个方面，提供了一种医疗智能分诊方法。所述方法可以包括：通过候选因素分析器从患者信息中抽取与疾病相关的症状和体征作为候选因素信息；鉴定知识挖掘器根据所述候选因素信息，从医学文献中获取与症状相关的多种候选疾病以及治疗措施作为鉴定知识信息；通过匹配器将所述鉴定知识信息与所述候选因素信息进行匹配；重复上述步骤，直到确定疾病所属科室或者所述患者信息已抽取并匹配完毕，返回疾病所属科室作为分诊结果。
在一个实施例中，所述患者信息可以通过人机交互接口由患者口头表达语言或电子录入获得，返回所确定科室作为分诊结果可以包括通过人机交互接口语音播报所确定科室或展示所确定科室的电子信息文本。
在一个实施例中，所述候选因素信息可以包括与疾病相关的症状和体征的关键词或者关键短语和时间事件，抽取所述候选因素信息的步骤可以包括：由所述候选因素分析器通过自然语言处理技术从患者提供的口头表达语言中或通过信息抽取技术从患者提供的信息文本中抽取所述关键词或者关键短语和时间事件。
在一个实施例中，获取所述鉴定知识信息的步骤可以包括：内容检索器根据所述候选因素信息中的关键词或者关键短语，从医学文献中检索与关键词或者关键短语相关的文档内容；知识抽取器利用自然语言处理技术从检索的文档内容中查找并挖掘与疾病相关的知识。
在一个实施例中，从医学文献中检索与关键词或者关键短语相关的文档内容的步骤可以包括：所述内容检索器使用自然语言处理技术对医学文献进行分词和命名实体识别，并建立倒排索引；以及所述内容检索器根据所述候选因素信息中的关键词或者关键短语，从倒排索引中检索出包含所有关键词或关键短语的文档。
在一个实施例中，利用自然语言处理技术从检索的文档内容中查 找并挖掘与疾病相关的知识的步骤可以包括：所述知识抽取器确定关键词或者关键短语在文档中所在的片段，并通过语义关系抽取与症状和体征相关的候选疾病。
在一个实施例中，将所述鉴定知识信息与所述候选因素信息进行匹配的步骤可以包括：如果输出的候选疾病只有一种，则根据所述候选因素确定所述疾病以及对应的所属科室；如果输出的候选疾病有多种，则根据具有相同的症状和体征的疾病在设定时间段或设定地域的流行度来确定所属科室；或者，进一步选择具有交叉的症状和体征的不同疾病之间区分度最大的信息作为延展问题，进一步获取患者信息并抽取新的候选因素信息以便进一步进行匹配过程。
在一个实施例中，可以通过信息增益来判断具有交叉的症状和体征的不同疾病之间的区分度，信息增益越大则区分度越大，信息增益越小则区分度越小，信息增益计算公式可以是：
IG(Symptom)＝H(Disease)-H(Disease|Symptom)
其中，Symptom表示症状，Disease表示疾病，H(.)表示熵。
根据本公开的另一个方面，提供了一种医疗智能分诊设备，包括候选因素分析器、鉴定知识挖掘器以及匹配器。所述候选因素分析器可以被配置用于从患者信息中抽取与疾病相关的症状和体征作为候选因素信息。所述鉴定知识挖掘器与所述候选因素分析器连接，可以被配置为根据所述候选因素信息，从医学文献中获取与症状相关的多种候选疾病以及治疗措施作为鉴定知识信息。所述匹配器与所述鉴定知识挖掘器相连，可以被配置用于将所述鉴定知识信息与所述候选因素信息进行匹配以便确定疾病所属科室，返回疾病所属科室作为分诊结果。
在一个实施例中，所述设备还可以包括人机交互设施，所述人机交互设施可以与所述候选因素分析器和所述匹配器分别连接。所述人机交互设施为患者提供人机交互接口，可以被配置用于收集患者信息以及向患者展示分诊结果。所述患者信息可以由患者口头表达语言或电子录入获得，所述分诊结果可以通过语音播报或电子信息文本展示返回至患者。
在一个实施例中，所述候选因素信息可以包括与疾病相关的症状和体征的关键词或者关键短语和时间事件，所述候选因素分析器可以被配置为通过自然语言处理技术从患者提供的口头表达语言中或通过信息抽取技术从患者提供的信息文本中，抽取所述关键词或者关键短语和时间事件。
在一个实施例中，所述鉴定知识挖掘器可以包含内容检索器和知识抽取器。所述内容检索器可以被配置为根据所述候选因素分析器输出的关键词或者关键短语，从医学文献中检索与关键词或者关键短语相关的文档内容。所述知识抽取器可以被配置为利用自然语言处理技术从检索的文档内容中查找并挖掘与疾病相关的知识。
在一个实施例中，所述内容检索器可以被配置用来：使用自然语言处理技术对医学文献进行分词和命名实体识别，并建立倒排索引；以及依据候选因素分析器输出的关键词或者关键短语，从倒排索引中检索出包含所有关键词或关键短语的文档。
在一个实施例中，所述知识抽取器可以被配置为确定关键词或者关键短语在文档中所在的片段，并通过语义关系抽取与症状和体征相关的候选疾病。
在一个实施例中，如果输出的候选疾病只有一种，则所述匹配器可以被配置用来根据所述候选因素确定所述疾病以及对应的所属科室。如果输出的候选疾病有多种，则所述匹配器可以被配置用来根据具有相同的症状和体征的疾病在设定时间段或设定地域的流行度来确定所属科室；或者，所述匹配器可以被配置用来进一步选择具有交叉的症状和体征的不同疾病之间区分度最大的信息作为延展问题，进一步获取患者信息并抽取新的候选因素信息以便进一步进行匹配过程。
在一个实施例中，所述匹配器可以被配置用来通过信息增益来判断具有交叉的症状和体征的不同疾病之间的区分度，信息增益越大则区分度越大，信息增益越小则区分度越小。信息增益计算公式可以是：
IG(Symptom)＝H(Disease)-H(Disease|Symptom)
其中，Symptom表示症状，Disease表示疾病，H(.)表示熵。
根据本公开的另一个方面，提供了一种医疗智能分诊设备，包括： 一个或多个处理器和存储器。该存储器上存储有计算机可执行指令，所述计算机可执行指令被配置为当被所述一个或多个处理器执行时，执行如上所述的任意一种方法。
根据本公开的另一个方面，提供了一种计算机可读存储介质，其上包含有计算机可执行指令，所述指令在由一个或多个处理器执行时，使所述一个或多个处理器执行如上所述的任意一种方法。
本公开的一些实施例可以实现如下有益效果中的至少一个有益效果和/或其它有益效果：该医疗智能分诊方法和医疗智能分诊设备通过分析和患者的交互内容，从医学文献或者其他医学数据中发现相关知识和规律从而形成鉴定知识信息的医学证据，进而分析患者的患病情况并给出相关建议。从而达到根据患者的主要症状及体征，匹配出可能的疾病及其隶属专科，并推荐有效的就诊的科室或就医路径的目的。
本发明内容部分以简化的形式介绍了本发明的一些构思，这些构思在下面的具体实施方式中进一步加以描述。本发明内容部分并非要给出要求保护的主题的必要特征或实质特征，也不是要限制要求保护的主题的范围。此外，正如本文所描述的，各种各样的其他特征和优点也可以根据需要结合到这些技术中。
附图说明
为了更清楚地说明本公开一些实施例的技术方案，本公开提供了下列附图以便在实施例描述时使用。应当意识到，下面描述中的附图仅仅涉及一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其它的附图，所述其它的附图也在本发明的范围内。
图1为根据一个实施例的医疗智能分诊方法的流程图；
图2为根据一个实施例的医疗智能分诊设备的架构示意图；
图3为根据一个实施例的分诊实例示意图；
附图标识：
1-人机交互设施；
2-候选因素分析器；
3-鉴定知识挖掘器；
4-匹配器；
5-患者；
6-医学文献。
具体实施方式
为使本领域技术人员更好地理解本公开的技术方案，下面结合附图和具体实施方式对本公开的医疗智能分诊方法、医疗智能分诊设备以及计算机可读介质作进一步详细描述。
本发明一些实施例的技术构思在于：医疗智能分诊是指根据患者的主要症状及体征，确定可能的疾病，判断病情的轻重缓急及其隶属专科，并推荐有效的就诊的路径。根据一些实施例的医疗智能分诊方法以及相应的医疗智能分诊设备基于知识挖掘(Knowledge Mining)和语义关系(Semantic Relation)，通过患者提供的信息从医学文献中挖掘相关的医学知识，自动地选择和患者与疾病相关的交互内容，从而更快更准地确定患者的分诊情况。
针对目前的分诊方式除了耽误病情诊断，往往还伴随着很高的时间成本和交通成本的问题，上述医疗智能分诊方法通过基于知识挖掘进行分诊，实现了自动地选择和患者的交互内容，从而更快更准地确定患者的分诊科室。
图1示出了根据一个实施例的医疗智能分诊方法的流程图。如图1所示，该医疗智能分诊方法可以包括步骤S1-S5。
步骤S1)通过人机交互接口收集患者信息。
在该步骤中，患者信息可以通过人机交互接口由患者口头表达语言或电子录入获得，分诊结果也可以利用人机交互接口通过语音播报或电子信息文本展示返回至患者。具体可以通过多种途径收集患者信息，例如：人机文字交互或语音交互，以获取患者包括主要症状的尽可能多的信息。
这里应该理解的是，患者信息可以一次性获取留待后续分析、匹配处理；也可以多次分步获取，逐步提供更多新的患者信息进行分析、匹配处理。
步骤S2)通过候选因素分析器从患者信息中抽取与疾病相关的症状和体征作为候选因素信息。
在该步骤中，候选因素分析器分析患者提供的信息，从中抽取和 疾病相关的主要症状和主要体征作为候选因素信息。在一个实施例中，可以通过自然语言处理技术从患者提供的口头表达语言中或通过信息抽取技术从患者提供的信息文本中，抽取与疾病相关的症状和体征的关键词或者关键短语和时间事件，上述关键词或者关键短语和时间事件可以作为候选因素信息。
步骤S3)鉴定知识挖掘器根据候选因素信息，从医学文献中获取与症状相关的多种候选疾病以及治疗措施作为鉴定知识信息。
在该步骤中，鉴定知识挖掘器可以从海量医学文献中挖掘与主要症状相关的疾病以及治疗措施的相关知识信息。即可以根据候选因素信息中的关键词或者关键短语，从医学文献中检索与关键词或者关键短语相关的文档内容；以及利用自然语言处理技术从检索的文档内容中查找并挖掘与疾病相关的知识。
从医学文献中获取与主要症状相关的疾病以及治疗措施等相关的鉴定知识信息的步骤可以进一步包括：
步骤S31)：内容检索器依据步骤S2)中输出的关键词或者关键短语，从医学文献中检索相关内容。即，内容检索器使用自然语言处理技术对医学文献进行分词和命名实体识别，并建立倒排索引；以及，依据步骤S2)中输出的关键词或者关键短语，从倒排索引中检索出包含所有关键词或关键短语的文档。
步骤S32)：知识抽取器利用自然语言处理技术从检索的文档中查找并挖掘与疾病相关知识。即，知识抽取器确定关键词或者关键短语在文档中所在的片段，并通过语义关系抽取和主要症状和主要体征相关的候选疾病。
步骤S4)通过匹配器将鉴定知识信息与候选因素信息进行匹配。
在该步骤中，匹配器将从步骤S3)获取的与主要症状相关的疾病以及治疗措施的相关知识信息中，抽取医学证据，并根据医学证据对候选因素信息进行匹配。所述匹配可能包括以下情况：
如果输出的候选疾病只有一种，则根据候选因素确定该候选疾病以及对应的所属科室；
如果输出的候选疾病有多种，则根据具有相同的症状和体征的疾病在设定时间段或设定地域的流行度确定所属科室；或者，进一步选择具有交叉的症状和体征的不同疾病之间区分度最大的信息作为延展 问题，进一步获取患者信息并抽取新的候选因素信息，以便进一步进行匹配过程。这里获取患者信息的方式可以向患者进一步提问咨询，目的即进一步获得与疾病相关的信息，并根据与疾病相关的候选因素和可能疾病的信息确定分诊结果，获得分诊结果。
在一个实施例中，可以通过信息增益来判断具有交叉的症状和体征的不同疾病之间的区分度，信息增益越大则区分度越大，信息增益越小则区分度越小，信息增益计算公式可以是：
IG(Symptom)＝H(Disease)-H(Disease|Symptom)
其中，Symptom表示症状，Disease表示疾病，H(.)表示熵；
可以通过进一步对多个候选疾病进行筛选和定位，获得实际最可能的分诊结果。
步骤S5)重复上述步骤，直到确定疾病所属科室或者患者信息已抽取并匹配完毕，返回疾病所属科室作为分诊结果。
重复步骤S1)-步骤S4)，直到确定疾病所属科室，或者不能再从患者信息中获得任何与疾病相关的信息，则返回疾病所属科室并向患者展示可能患病的分诊结果。
通过上述步骤的往复交互和分析，直到自动返回分诊结果。该分诊结果从患者信息出发抽取与疾病相关的症状和体征作为候选因素信息，并从海量医学文献中自动查找并挖掘与疾病相关的鉴定知识信息，因此分诊结果最接近实际的应挂号科室。
根据本公开的另一个方面，还提供了一种医疗智能分诊设备。同样基于知识挖掘进行分诊，该分诊设备能够自动地选择和患者信息中与疾病相关的候选因素信息，从而更快更准地确定患者的分诊科室。
图2示出了根据一个实施例的医疗智能分诊设备的架构示意图。如图2所示，该医疗智能分诊设备可以包括人机交互设施1、候选因素分析器2、鉴定知识挖掘器3以及匹配器4。
人机交互设施1可以为患者5提供人机交互接口，可以被配置用于收集患者信息以及向患者5展示分诊结果。其中，患者信息可以由患者口头表达语言或电子录入获得，分诊结果可以通过语音播报或电子信息文本展示返回至患者。
候选因素分析器2与人机交互设施1连接，可以被配置用于从患者信息中抽取与疾病相关的症状和体征作为候选因素信息。例如，候选因素分析器2可以被配置用于分析人机交互设施1收集的患者5提供的信息，从中抽取和疾病相关的主要症状和主要体征作为候选因素信息。在一个实施例中，候选因素分析器2可以被配置为通过自然语言处理技术从患者提供的口头表达语言中或通过信息抽取技术从患者提供的信息文本中，抽取与疾病相关的症状和体征的关键词或者关键短语和时间事件，上述关键词或者关键短语和时间事件可以作为候选因素信息。通过候选因素分析器2提取患者信息中与疾病相关的信息，可以获得疾病候选因素。
鉴定知识挖掘器3与候选因素分析器2连接，可以被配置为根据候选因素信息，从医学文献中获取与症状相关的多种候选疾病以及治疗措施作为鉴定知识信息。可以根据从候选因素分析器2抽取的信息，从海量医学文献6中挖掘与主要症状相关的疾病以及治疗措施的相关知识信息，获得与疾病候选因素对应的可能疾病的信息。
在一个实施例中，鉴定知识挖掘器3可以包含内容检索器和知识抽取器。内容检索器可以被配置为根据候选因素分析器输出的关键词或者关键短语，从医学文献中检索与关键词或者关键短语相关的文档内容。知识抽取器可以被配置为利用自然语言处理技术从检索的文档内容中查找并挖掘与疾病相关的知识。
内容检索器可以被配置为使用自然语言处理技术对医学文献进行分词和命名实体识别，并建立倒排索引；以及，依据候选因素分析器输出的关键词或者关键短语，从倒排索引中检索出包含所有关键词或关键短语的文档。知识抽取器可以被配置为确定关键词或者关键短语在文档中所在的片段，并通过语义关系抽取与症状和体征相关的候选疾病。
匹配器4可以分别与人机交互设施1和鉴定知识挖掘器3相连，可以被配置用于将鉴定知识信息与候选因素信息进行匹配，返回分诊结果。可以通过从鉴定知识挖掘器3输出的知识和候选因素信息的比对，匹配患者5可能患的疾病的分诊结果。如果输出的候选疾病只有一种，则匹配器4可以被配置用来根据候选因素确定该候选疾病以及对应的所属科室。如果输出的候选疾病有多种，则匹配器4可以被配 置用来根据具有相同的症状和体征的疾病在设定时间段或设定地域的流行度来确定所属科室；或者，进一步选择具有交叉的症状和体征的不同疾病之间区分度最大的信息作为延展问题，进一步获取患者信息并抽取新的候选因素信息，以便进一步进行匹配过程。
匹配器4可以通过信息增益来判断具有交叉的症状和体征的不同疾病之间的区分度，信息增益越大则区分度越大，信息增益越小则区分度越小。信息增益计算公式可以是：
IG(Symptom)＝H(Disease)-H(Disease|Symptom)
其中，Symptom表示症状，Disease表示疾病，H(.)表示熵。
通过匹配器4进一步对多个候选疾病进行筛选和定位，可以获得实际最可能的分诊结果。
根据本公开的另一个方面，医疗智能分诊设备还可以通过一个或多个处理器和存储器来实现。该存储器上存储有计算机可执行指令，所述计算机可执行指令被配置为当被所述一个或多个处理器执行时，执行如上所述的任意一种方法。
根据本公开的另一个方面，提供了一种计算机可读存储介质，其上包含有计算机可执行指令，所述指令在由一个或多个处理器执行时，使所述一个或多个处理器执行如上所述的任意一种方法。
图3示出了根据一个实施例的分诊实例示意图。下面将参考图3结合一个分诊实例对上述医疗智能分诊方法和医疗智能分诊设备各模块的功能和实现方式进行详细的说明。
需要指出的是，医疗智能分诊设备中的四大模块与医疗智能分诊方法中的五大步骤的对应关系为：人机交互设施1涉及步骤S 1)和步骤S5)、候选因素分析器2涉及步骤S2)、鉴定知识挖掘器3涉及步骤S3)以及匹配器4涉及步骤S4)。
首先，通过人机交互设施1为患者5提供人机交互接口，可以被配置用于收集患者5提供的症状和体征等信息。例如，可以通过询问患者5不舒服的部位或者提供相应的症状、体征等，然后收集患者5的回复信息形成患者信息。人机交互设施1还可以被配置用于向患者5展示该医疗智能分诊设备最终的分诊结果。
接着，通过候选因素分析器2分析人机交互设施1收集的患者信息，从中抽取和疾病相关的症状和体征等主要信息。在一个实施例中，可以通过自然语言处理技术和信息抽取技术从患者信息语音或文本中抽取关键词或者关键短语(这里的关键词或关键短语指和疾病相关的症状体征等)和时间事件。例如，设备问“请问您哪里不舒服？”，患者5可以回答“昨天开始肚子痛，今天腰也开始痛了”。从这个例子中，候选因素分析器2可以自动抽取出关键词“肚子痛”和关键短语“腰也开始痛了”。同时，可以抽取时间名词“昨天”和“今天”，进而分析得到时间、事件(昨天、肚子痛)和(今天、肚子痛和腰痛)。
然后，通过鉴定知识挖掘器3依据候选因素分析器2抽取的信息，从海量医学文献6中挖掘相关知识信息，从前述内容可知：鉴定知识挖掘器3可以包含内容检索器和知识抽取器。
内容检索器可以依据候选因素分析器2输出的关键词或者关键短语，从医学文献6中检索相关内容。在一个实施例中可以包括如下过程：先是使用自然语言处理技术对医学文献6进行分词和命名实体识别，并建立倒排索引，所述倒排索引可以采取如下形式：“词语1”-“文档1，文档i，....，文档N”，其中“文档1，文档i，文档N”都是包含“词语1”的文档；再依据候选因素分析器2输出的关键词或者关键短语，从倒排索引中检索出包含所有关键词或关键短语的文档。“w1，w2，...，wK”为K个关键词或者关键短语，“S1，S2，...，SK”为倒排索引中对应的文档集合(即S1为包含w1的文档集合，依此类推)，则检索出的文档集为S＝{S1∩S2∩...∩SK}。例如，假设关键词为“头疼”和“眩晕”，在倒排索引中“头疼”对应的文档集为{“文档1”、”文档2”、”文档3”}，“眩晕”对应的文档集为{“文档2”，”文档4”，”文档6”}，则检索出符合条件的文档集为{“文档2”}。
知识抽取器可以利用自然语言处理技术从检索的文档中挖掘相关知识。在一个实施例中可以包括如下过程：确定关键词或者关键短语在文档中所在的片段，在通过语义关系抽取和症状体征(关键词或者关键短语，如头疼、眩晕等)相关的候选疾病。例如，给定关键词“头疼”和“眩晕”，从检索的文档中确定片段“高血压会引起动脉充血、扩张，产生头疼，甚至有时引发恶心、呕吐、眩晕。”。通过语义关系抽取，知识抽取器可以获得如下的语义关系：高血压引发头疼、高血 压引发恶心、高血压引发呕吐、高血压引发眩晕等等。
通过上述的语义关系，该内容检索器挖掘出知识{引发(高血压，头疼)，引发(高血压，恶心)，引发(高血压，呕吐)，引发(高血压，眩晕)}，进而将“高血压”作为引起患者5头疼眩晕的候选疾病。
最后，通过匹配器4从鉴定知识挖掘器3输出的知识和候选因素信息中分析患者5所患的可能疾病的分诊情况。在一个实施例中，所述分析可以包括如下过程：
如果鉴定知识挖掘器3输出的候选疾病只有一种，则输出该疾病以及对应的科室。例如，如果针对“头疼”和“眩晕”，鉴定知识挖掘器3输出的候选疾病只有“高血压”，则输出“高血压”以及应挂号的科室“心血管内科”；
如果输出的候选疾病有多种，则可以选择区分度最大的信息作为问题向患者5咨询。具体来说，通过信息增益来判断区分度，信息增益越大则区分度越大，信息增益越小则区分度越小。例如，针对上例，如果输出的候选疾病有“高血压”、“偏头痛”以及“神经衰弱”，返回的知识如表1所示：
表1 返回知识表
疾病 关系 症状或体征
高血压 引发 头疼、眩晕、恶心、尿多
偏头痛 引发 头疼、眩晕、恶心、呕吐
神经衰弱 引发 头疼、眩晕、失眠、焦虑、烦躁
从表1中可以看出，仅仅依靠“头疼”和“眩晕”无法判断患者5可能患有的疾病，该疾病可能为高血压、偏头痛或神经衰弱中的任一种。因此，根据新知识进一步的交互提问，设备需要通过人机交互器1和患者5交流，收集新的患者信息来确定可能患有的疾病症状或体征。针对上例，设备可以提出问题“您有恶心的症状吗？”或者“您最近失眠吗？”或者其他任何相关症状的问题。由于病种和症状通常很多，因此设备如何选择症状以便尽快确定患者5的患病信息变得至关重要。
通过进一步对多个候选疾病信息进行筛选和定位，获得实际最可能的分诊结果。该医疗智能分诊方法和医疗智能分诊设备从两个方面 考虑选择症状：
1)疾病的流行度。即医院已经确诊的具有相同症状的病人所患疾病所占的比例。例如假设具有“头疼、眩晕”症状的100人中，70人确诊为高血压，20人确诊为偏头痛，10人为神经衰弱，则高血压的流行度为popular(高血压)＝70％，popular(偏头痛)＝20％，popular(神经衰弱)＝10％。本实施例的医疗智能分诊方法将疾病的流行度看作先验概率，在其他条件相同的情况下，优先选择流行度大的疾病的症状。比如，在上述疾病流行度比例条件下，匹配表1中可能的疾病为高血压；
2)症状的区分度。即症状是否对疾病具有区分性。例如，针对上例，如果设备选择问“您有恶心的症状吗？”，如果患者5回答“是”，设备仍然无法判断患者5是患有高血压还是患有偏头痛，因为这两种疾病都能引起恶心的症状，可能还需要进一步获得新的候选因素。
本实施例的医疗智能分诊方法采用信息增益公式来计算症状的区分度，即：IG(Symptom)＝H(Disease)-H(Disease|Symptom)，其中，Symptom表示症状，Disease表示疾病，H(.)表示熵。
根据表1信息，假设一个具有“头疼、眩晕”症状的患者来分诊，此时设备无法确定患者是该分诊为“高血压”，“偏头痛”还是该分诊为“神经衰弱”，因为这三种疾病都可能引起上述症状。为了进一步明确患者的情况，设备需要收集更多的患者信息。
设备首先根据疾病的流行度判断具有上述症状的哪种疾病的患者最多，例如在本示例中假设高血压的患者所占比例为70％(可以通过统计医院确诊的病历来获取该百分比)，即大部分情况下出现“头疼、眩晕”症状的患者是患有高血压。此时设备会从鉴定知识挖掘器3得到的高血压相关症状中选择一个症状来询问患者以便确定患者是否患有高血压的可能。
其次，除去“头疼、眩晕”，高血压还有两个相关症状“恶心”和“尿多”。对于这两个症状，设备分别计算它们的信息增益(IG)，并选择信息增益最大的症状向患者获取相关的信息。假设“高血压”，“偏头痛”和“神经衰弱”服从均匀分布，则p(高血压)＝p(偏头痛)＝p(神经衰弱)＝1/3，其中p(.)表示概率。此时与“头疼、眩晕”症状相关疾病的熵为：
Figure PCTCN2017116370-appb-000001
由于在本示例中出现“恶心”的疾病只有高血压和偏头痛，当症状为“恶心”的时候，p(高血压)＝1/2，p(偏头痛)＝1/2，p(神经衰弱)＝0，此时与“恶心”症状相关疾病的熵为：
Figure PCTCN2017116370-appb-000002
则根据症状的区分度的信息增益公式，有：
Figure PCTCN2017116370-appb-000003
由于在本示例中出现“尿多”的疾病只有高血压，因此当症状为“尿多”的时候，p(高血压)＝1，p(偏头痛)＝0，p(神经衰弱)＝0，此时与“尿多”症状相关疾病的熵为：
Figure PCTCN2017116370-appb-000004
Figure PCTCN2017116370-appb-000005
因为IG(尿多)＞IG(恶心)，所以设备选择症状“尿多”向患者进一步获取信息或提问(例如，您最近出现尿多情况吗？)。如果患者提供的信息或回答为“是”，则确定为高血压分诊结果。如果回答“否”，则排除高血压分诊结果，并将剩下的疾病作为候选疾病。可见，对于本示例而言，分诊匹配是对“偏头痛”和“神经衰弱”重复上述步骤，直到确定疾病分诊或者患者终止该过程。
综上所述，本实施例的医疗智能分诊方法及其相应的医疗智能分诊设备，首先通过疾病的流行度，选择流行度最大的疾病，然后从该疾病中选择最具有区分度的症状作为问题，通过人机交互器1向患者5询问。例如，对于上例，首先选择高血压作为患者5最有可能患有的 疾病(因为高血压的流行度最大70％)，然后从知识抽取器输出的知识中，寻找高血压的其他症状，选择区分度最大的症状“尿多”，产生问题“您最近尿多吗？”通过人机交互器1向患者5进行询问。如此往复循环直到确定疾病分诊或者不能再获取新的疾病候选因素信息，此时将设备匹配的分诊结果通过人机交互器1展现给用户。
该医疗智能分诊方法和医疗智能分诊设备通过分析和患者的交互内容，从医学文献或者其他医学数据中发现相关知识和规律从而形成鉴定知识信息的医学证据，进而分析患者的患病情况并给出相关建议。从而达到根据患者的主要症状及体征，匹配出可能的疾病及其隶属专科，并推荐有效的就诊的科室或就医路径的目的。
上述基于知识挖掘的医疗智能分诊方法和医疗智能分诊设备有如下两大优点：
1)通过分析患者提供的信息以及相关的医学知识，能够自动地选择和患者的交互内容，从而更快更准地确定患者的患病所属科室情况；
2)通过知识挖掘确定相关证据，这些证据使用自然语言描述，具有可解释性。然后，依据支持度列出患者所有可能的情况，并给出合理的建议。
可以理解的是，以上实施方式仅仅是为了说明本发明的原理而采用的示例性实施方式，然而本发明并不局限于此。应当指出的是，对于本领域内的普通技术人员而言，在不脱离本发明的精神和实质的情况下，可以做出各种变型和改进，这些变型和改进也被涵盖在本发明的保护范围之内。本发明的保护范围应以所附权利要求的保护范围为准。
需要说明的是，上述实施例仅以上述各功能模块的划分进行举例说明，实际应用中，可以根据需要将上述功能分配给不同的功能模块完成。可以将装置的内部结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。另外，上述一个模块的功能可以由多个模块来完成，上述多个模块的功能也可以集成到一个模块中完成。
在权利要求书中，任何置于括号中的附图标记都不应当解释为限制权利要求。术语“包括”并不排除除了权利要求中所列出的元件或步骤之外的元件或步骤的存在。元件前的词语“一”或“一个”并不排除存在多个这样的元件。本发明可以借助于包括若干分离元件的硬 件来实现，也可以通过适当编程的软件或固件来实现，或者通过它们的任意组合来实现。
在列举了若干装置的设备或系统权利要求中，这些装置中的一个或多个能够在同一个硬件项目中体现。仅仅某些措施记载在相互不同的从属权利要求中这个事实并不表明这些措施的组合不能被有利地使用。

Claims (18)

  1. 一种医疗智能分诊方法，所述方法包括：
    通过候选因素分析器从患者信息中抽取与疾病相关的症状和体征作为候选因素信息；
    鉴定知识挖掘器根据所述候选因素信息，从医学文献中获取与症状相关的多种候选疾病以及治疗措施作为鉴定知识信息；
    通过匹配器将所述鉴定知识信息与所述候选因素信息进行匹配；
    重复上述步骤，直到确定疾病所属科室或者所述患者信息已抽取并匹配完毕，返回疾病所属科室作为分诊结果。
  2. 根据权利要求1所述的医疗智能分诊方法，其中，所述患者信息通过人机交互接口由患者口头表达语言或电子录入获得，返回所确定科室作为分诊结果包括通过人机交互接口语音播报所确定科室或展示所确定科室的电子信息文本。
  3. 根据权利要求1所述的医疗智能分诊方法，其中，所述候选因素信息包括与疾病相关的症状和体征的关键词或者关键短语和时间事件，抽取所述候选因素信息的步骤包括：由所述候选因素分析器通过自然语言处理技术从患者提供的口头表达语言中或通过信息抽取技术从患者提供的信息文本中抽取所述关键词或者关键短语和时间事件。
  4. 根据权利要求3所述的医疗智能分诊方法，其中，获取所述鉴定知识信息的步骤包括：
    内容检索器根据所述候选因素信息中的关键词或者关键短语，从医学文献中检索与关键词或者关键短语相关的文档内容；
    知识抽取器利用自然语言处理技术从检索的文档内容中查找并挖掘与疾病相关的知识。
  5. 根据权利要求4所述的医疗智能分诊方法，其中，从医学文献中检索与关键词或者关键短语相关的文档内容的步骤包括：
    所述内容检索器使用自然语言处理技术对医学文献进行分词和命名实体识别，并建立倒排索引；以及，
    所述内容检索器根据所述候选因素信息中的关键词或者关键短语，从倒排索引中检索出包含所有关键词或关键短语的文档。
  6. 根据权利要求5所述的医疗智能分诊方法，其中，利用自然语 言处理技术从检索的文档内容中查找并挖掘与疾病相关的知识的步骤包括：所述知识抽取器确定关键词或者关键短语在文档中所在的片段，并通过语义关系抽取与症状和体征相关的候选疾病。
  7. 根据权利要求1所述的医疗智能分诊方法，其中，将所述鉴定知识信息与所述候选因素信息进行匹配的步骤包括：
    如果输出的所述候选疾病只有一种，则根据所述候选因素确定所述疾病以及对应的所属科室；
    如果输出的所述候选疾病有多种，则根据具有相同的症状和体征的疾病在设定时间段或设定地域的流行度来确定所属科室；或者，进一步选择具有交叉的症状和体征的不同疾病之间区分度最大的信息作为延展问题，进一步获取患者信息并抽取新的候选因素信息以便进一步进行匹配过程。
  8. 根据权利要求7所述的医疗智能分诊方法，其中，通过信息增益来判断具有交叉的症状和体征的不同疾病之间的区分度，信息增益越大则区分度越大，信息增益越小则区分度越小，信息增益计算公式为：
    IG(Symptom)＝H(Disease)-H(Disease|Symptom)
    其中，Symptom表示症状，Disease表示疾病，H(.)表示熵。
  9. 一种医疗智能分诊设备，包括候选因素分析器、鉴定知识挖掘器以及匹配器，其中：
    所述候选因素分析器被配置用于从患者信息中抽取与疾病相关的症状和体征作为候选因素信息；
    所述鉴定知识挖掘器与所述候选因素分析器连接，被配置为根据所述候选因素信息，从医学文献中获取与症状相关的多种候选疾病以及治疗措施作为鉴定知识信息；
    所述匹配器与所述鉴定知识挖掘器相连，被配置用于将所述鉴定知识信息与所述候选因素信息进行匹配以便确定疾病所属科室，返回所确定科室作为分诊结果。
  10. 根据权利要求9所述的医疗智能分诊设备，其中，所述设备还包括人机交互设施，所述人机交互设施与所述候选因素分析器和所述 匹配器分别连接，所述人机交互设施为患者提供人机交互接口，被配置用于收集患者信息以及向患者展示分诊结果；其中，所述患者信息由患者口头表达语言或电子录入获得，所述分诊结果通过语音播报或电子信息文本展示返回至患者。
  11. 根据权利要求9所述的医疗智能分诊设备，其中，所述候选因素信息包括与疾病相关的症状和体征的关键词或者关键短语和时间事件，所述候选因素分析器被配置为通过自然语言处理技术从患者提供的口头表达语言中或通过信息抽取技术从患者提供的信息文本中，抽取所述关键词或者关键短语和时间事件。
  12. 根据权利要求11所述的医疗智能分诊设备，其中，所述鉴定知识挖掘器包含内容检索器和知识抽取器，其中：
    所述内容检索器被配置为根据所述候选因素分析器输出的关键词或者关键短语，从医学文献中检索与关键词或者关键短语相关的文档内容；
    所述知识抽取器被配置为利用自然语言处理技术从检索的文档内容中查找并挖掘与疾病相关的知识。
  13. 根据权利要求12所述的医疗智能分诊设备，其中，所述内容检索器被配置用来：
    使用自然语言处理技术对医学文献进行分词和命名实体识别，并建立倒排索引；以及
    依据候选因素分析器输出的关键词或者关键短语，从倒排索引中检索出包含所有关键词或关键短语的文档。
  14. 根据权利要求13所述的医疗智能分诊设备，其中，所述知识抽取器被配置为确定关键词或者关键短语在文档中所在的片段，并通过语义关系抽取与症状和体征相关的候选疾病。
  15. 根据权利要求9所述的医疗智能分诊设备，其中，如果输出的所述候选疾病只有一种，则所述匹配器被配置用来根据所述候选因素确定所述疾病以及对应的所属科室；
    如果输出的所述候选疾病有多种，则所述匹配器被配置用来根据具有相同的症状和体征的疾病在设定时间段或设定地域的流行度来确定所属科室；或者，所述匹配器被配置用来进一步选择具有交叉的症状和体征的不同疾病之间区分度最大的信息作为延展问题，进一步获 取患者信息并抽取新的候选因素信息以便进一步进行匹配过程。
  16. 根据权利要求15所述的医疗智能分诊设备，其中，所述匹配器被配置用来通过信息增益来判断具有交叉的症状和体征的不同疾病之间的区分度，信息增益越大则区分度越大，信息增益越小则区分度越小，信息增益计算公式为：
    IG(Symptom)＝H(Disease)-H(Disease|Symptom)
    其中，Symptom表示症状，Disease表示疾病，H(.)表示熵。
  17. 一种医疗智能分诊设备，包括：
    一个或多个处理器；和
    存储器，其上存储有计算机可执行指令，所述计算机可执行指令被配置为当被所述一个或多个处理器执行时，执行如权利要求1-8中任何一项所述的方法。
  18. 一种计算机可读存储介质，其上包含有计算机可执行指令，所述指令在由一个或多个处理器执行时，使所述一个或多个处理器执行如权利要求1-8中任何一项所述的方法。
PCT/CN2017/116370 2017-05-12 2017-12-15 医疗智能分诊方法和设备 WO2018205609A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP17908895.0A EP3623970A4 (en) 2017-05-12 2017-12-15 MEDICAL INTELLIGENT SORTING PROCESS AND DEVICE
US16/069,139 US20210280301A1 (en) 2017-05-12 2017-12-15 Intelligent triage method and device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710333821.0A CN108877921B (zh) 2017-05-12 2017-05-12 医疗智能分诊方法和医疗智能分诊系统
CN201710333821.0 2017-05-12

Publications (1)

Publication Number Publication Date
WO2018205609A1 true WO2018205609A1 (zh) 2018-11-15

Family

ID=64104391

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/116370 WO2018205609A1 (zh) 2017-05-12 2017-12-15 医疗智能分诊方法和设备

Country Status (4)

Country Link
US (1) US20210280301A1 (zh)
EP (1) EP3623970A4 (zh)
CN (1) CN108877921B (zh)
WO (1) WO2018205609A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111128376A (zh) * 2019-11-21 2020-05-08 泰康保险集团股份有限公司 一种推荐评估表单的方法和装置
CN111326240A (zh) * 2019-07-15 2020-06-23 郑州大学第一附属医院 一种基于无模型的智慧医疗分诊推荐方法
CN111615697A (zh) * 2018-12-24 2020-09-01 北京嘀嘀无限科技发展有限公司 基于文本片段搜索的人工智能医学症状识别系统
CN112199509A (zh) * 2020-09-14 2021-01-08 山东众阳健康科技集团有限公司 一种基于知识图谱的导诊方法、系统和存储介质
CN112700865A (zh) * 2021-01-07 2021-04-23 重庆中肾网络科技有限公司 一种基于综合推理的智能分诊方法

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109659013B (zh) * 2018-11-28 2023-07-07 平安科技（深圳）有限公司 病症分诊及路径优化方法、装置、设备及存储介质
CN111276259B (zh) * 2018-12-04 2024-03-01 阿里巴巴集团控股有限公司 服务确定、网络交互、分类方法和客户端、服务器和介质
CN110415776A (zh) * 2019-07-26 2019-11-05 深圳市赛为智能股份有限公司 医疗管理方法、装置、计算机设备及存储介质
CN112837813A (zh) * 2019-11-25 2021-05-25 北京搜狗科技发展有限公司 自动问诊方法及装置
CN111383728A (zh) * 2020-02-24 2020-07-07 华中科技大学同济医学院附属同济医院 用于新冠肺炎隔离管理的医学症状信息处理装置及隔离管理系统
CN111462909B (zh) * 2020-03-30 2024-04-05 讯飞医疗科技股份有限公司 疾病演化跟踪和病情提示方法、装置及电子设备
CN111813957A (zh) * 2020-07-14 2020-10-23 深圳中兴网信科技有限公司 基于知识图谱的医疗导诊方法和可读存储介质
KR102241399B1 (ko) * 2020-08-25 2021-04-16 주식회사 쓰리빌리언 증상의 질병 특이도 측정 시스템
CN111985246B (zh) * 2020-08-27 2023-08-15 武汉东湖大数据交易中心股份有限公司 一种基于主要症状与伴随症状词的疾病认知系统
CN112016318B (zh) * 2020-09-08 2023-11-21 平安科技（深圳）有限公司 基于解释模型的分诊信息推荐方法、装置、设备及介质
CN113571167B (zh) * 2021-07-28 2024-04-19 重庆橡树信息科技有限公司 一种基于配置式评分知识模型的快捷分诊系统
CN116665865B (zh) * 2023-06-13 2023-12-26 爱汇葆力(广州)数据科技有限公司 基于大数据实现陪诊人员的信息智能管理方法及系统
CN116759078B (zh) * 2023-08-21 2023-12-08 药融云数字科技(成都)有限公司 支持双语输入的疾病循证方法、系统、存储介质及终端
CN118016263B (zh) * 2024-04-09 2024-06-07 广州市挖米科技有限责任公司 一种基于语音识别的数字化医疗助手系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184315A (zh) * 2011-04-02 2011-09-14 中国医学科学院医学信息研究所 基于诊断要素分析的科室分诊系统
CN103164616A (zh) * 2013-02-02 2013-06-19 杭州卓健信息科技有限公司 一种智能导诊系统和方法
CN103870673A (zh) * 2013-09-03 2014-06-18 北京天鹏恒宇科技发展有限公司 医疗研发系统使用支持文档的结构化标记法
CN106295186A (zh) * 2016-08-11 2017-01-04 中国科学院计算技术研究所 一种基于智能推理的辅助疾病诊断的方法与系统

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6915254B1 (en) * 1998-07-30 2005-07-05 A-Life Medical, Inc. Automatically assigning medical codes using natural language processing
US7379885B1 (en) * 2000-03-10 2008-05-27 David S. Zakim System and method for obtaining, processing and evaluating patient information for diagnosing disease and selecting treatment
US20030105638A1 (en) * 2001-11-27 2003-06-05 Taira Rick K. Method and system for creating computer-understandable structured medical data from natural language reports
US20090187425A1 (en) * 2007-09-17 2009-07-23 Arthur Solomon Thompson PDA software robots leveraging past history in seconds with software robots
CN101441636A (zh) * 2007-11-21 2009-05-27 中国科学院自动化研究所 一种基于知识库的医院信息搜索引擎及系统
US9521961B2 (en) * 2007-11-26 2016-12-20 C. R. Bard, Inc. Systems and methods for guiding a medical instrument
US20090240702A1 (en) * 2008-03-21 2009-09-24 Computerized Screening, Inc. Community based managed health kiosk and research database system
CN102129526A (zh) * 2011-04-02 2011-07-20 中国医学科学院医学信息研究所 面向公众的就医向导式自助分诊挂号方法及系统
US20140122109A1 (en) * 2012-10-29 2014-05-01 Consuli, Inc. Clinical diagnosis objects interaction
US20150161331A1 (en) * 2013-12-04 2015-06-11 Mark Oleynik Computational medical treatment plan method and system with mass medical analysis
CN104376409A (zh) * 2014-11-07 2015-02-25 深圳市前海安测信息技术有限公司 一种基于网络医院的分诊数据处理方法及系统
US10770184B1 (en) * 2014-12-04 2020-09-08 Cerner Innovation, Inc. Determining patient condition from unstructured text data
CN206021239U (zh) * 2016-06-27 2017-03-15 好人生（上海）健康科技有限公司 专门适用于医学预分诊的自然语言交互装置
CN106407721B (zh) * 2016-11-10 2019-05-07 上海电机学院 一种可实现分级诊疗的医生预约挂号方法
CN106557653B (zh) * 2016-11-15 2017-09-22 合肥工业大学 一种移动医疗智能导医系统及其方法
CN106650261A (zh) * 2016-12-22 2017-05-10 上海智臻智能网络科技股份有限公司 智能问诊方法、装置和系统
US20180218126A1 (en) * 2017-01-31 2018-08-02 Pager, Inc. Determining Patient Symptoms and Medical Recommendations Based on Medical Information
US10599771B2 (en) * 2017-04-10 2020-03-24 International Business Machines Corporation Negation scope analysis for negation detection

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184315A (zh) * 2011-04-02 2011-09-14 中国医学科学院医学信息研究所 基于诊断要素分析的科室分诊系统
CN103164616A (zh) * 2013-02-02 2013-06-19 杭州卓健信息科技有限公司 一种智能导诊系统和方法
CN103870673A (zh) * 2013-09-03 2014-06-18 北京天鹏恒宇科技发展有限公司 医疗研发系统使用支持文档的结构化标记法
CN106295186A (zh) * 2016-08-11 2017-01-04 中国科学院计算技术研究所 一种基于智能推理的辅助疾病诊断的方法与系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3623970A4 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111615697A (zh) * 2018-12-24 2020-09-01 北京嘀嘀无限科技发展有限公司 基于文本片段搜索的人工智能医学症状识别系统
CN111326240A (zh) * 2019-07-15 2020-06-23 郑州大学第一附属医院 一种基于无模型的智慧医疗分诊推荐方法
CN111326240B (zh) * 2019-07-15 2022-05-27 郑州大学第一附属医院 一种基于无模型的智慧医疗分诊推荐方法
CN111128376A (zh) * 2019-11-21 2020-05-08 泰康保险集团股份有限公司 一种推荐评估表单的方法和装置
CN111128376B (zh) * 2019-11-21 2023-06-16 泰康保险集团股份有限公司 一种推荐评估表单的方法和装置
CN112199509A (zh) * 2020-09-14 2021-01-08 山东众阳健康科技集团有限公司 一种基于知识图谱的导诊方法、系统和存储介质
CN112700865A (zh) * 2021-01-07 2021-04-23 重庆中肾网络科技有限公司 一种基于综合推理的智能分诊方法

Also Published As

Publication number Publication date
EP3623970A4 (en) 2021-01-13
US20210280301A1 (en) 2021-09-09
CN108877921A (zh) 2018-11-23
EP3623970A1 (en) 2020-03-18
CN108877921B (zh) 2021-10-19

Similar Documents

Publication Publication Date Title
WO2018205609A1 (zh) 医疗智能分诊方法和设备
US20190035506A1 (en) Intelligent auxiliary diagnosis method, system and machine-readable medium thereof
CN110085307B (zh) 一种基于多源知识图谱融合的智能导诊方法和系统
CN111316281B (zh) 基于机器学习的自然语言情境中数值数据的语义分类方法以及系统
CN106407443B (zh) 一种结构化医疗数据生成方法及装置
CN104699741B (zh) 用于改善对输入问题的回答的方法和装置
CN105808931B (zh) 一种基于知识图谱的针灸决策支持方法及装置
CN112786194A (zh) 基于人工智能的医学影像导诊导检系统、方法及设备
US8949079B2 (en) Patient data mining
CN111414393A (zh) 一种基于医学知识图谱的语义相似病例检索方法及设备
CN111813957A (zh) 基于知识图谱的医疗导诊方法和可读存储介质
CN110675944A (zh) 分诊方法及装置、计算机设备及介质
US20200075135A1 (en) Trial planning support apparatus, trial planning support method, and storage medium
CN112232065A (zh) 挖掘同义词的方法及装置
CN111564223A (zh) 传染病生存概率的预测方法、预测模型的训练方法及装置
CN114611735A (zh) 就医的互联网挂号方法、装置、设备及存储介质
CN113571184A (zh) 一种用于精神健康测评的对话交互设计方法及系统
CN112634889A (zh) 基于人工智能的电子病例录入方法、装置、终端及介质
CN113488157B (zh) 智能导诊处理方法、装置、电子设备及存储介质
CN110377698B (zh) 基于阅读理解的任务处理方法及装置、设备及可读介质
CN107610783A (zh) 一种脑肿瘤信息平台
CN109522331B (zh) 以个人为中心的区域化多维度健康数据处理方法及介质
CN110993116A (zh) 诊疗数据提取方法及装置、电子设备、存储介质
Norman Systematic review automation methods
CN112116976A (zh) 药品信息的处理方法、装置及计算机可读存储介质

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: 17908895

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 2017908895

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2017908895

Country of ref document: EP

Effective date: 20191212