CN112216383B - Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning - Google Patents

Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning Download PDF

Info

Publication number
CN112216383B
CN112216383B CN202011155816.3A CN202011155816A CN112216383B CN 112216383 B CN112216383 B CN 112216383B CN 202011155816 A CN202011155816 A CN 202011155816A CN 112216383 B CN112216383 B CN 112216383B
Authority
CN
China
Prior art keywords
syndrome
disease
inquiry
symptom
candidate
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN202011155816.3A
Other languages
Chinese (zh)
Other versions
CN112216383A (en
Inventor
孙钊
吴军
王爱
高希余
樊昭磊
冯德杰
王琪琪
李涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhongyang Health Technology Group Co ltd
Original Assignee
Shandong Msunhealth Technology Group Co Ltd
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 Shandong Msunhealth Technology Group Co Ltd filed Critical Shandong Msunhealth Technology Group Co Ltd
Priority to CN202011155816.3A priority Critical patent/CN112216383B/en
Publication of CN112216383A publication Critical patent/CN112216383A/en
Application granted granted Critical
Publication of CN112216383B publication Critical patent/CN112216383B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20ICT 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
    • 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
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367Ontology
    • 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
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70ICT 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
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02ATECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Biomedical Technology (AREA)
  • Epidemiology (AREA)
  • Theoretical Computer Science (AREA)
  • Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Physics & Mathematics (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The invention provides a traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome and deep learning, which is used for analyzing symptom information input by a patient according to a preset knowledge graph based on the syndrome of symptoms; determining whether an inquiry needs to be made based on the result of the symptom analysis; if the judgment result is that the inquiry is needed, the traditional Chinese medicine inquiry is carried out by utilizing the syndrome and the deep learning network model, and the syndrome and/or the syndrome of the patient are/is obtained according to the answer of the patient; judging whether tongue diagnosis is combined, if so, entering the result of the inquiry module into a combined tongue diagnosis module to obtain the final syndrome and/or syndrome element; otherwise, the result obtained by the inquiry is taken as the final syndrome and/or syndrome element; when the inquiry judging module judges that inquiry is not needed or the tongue diagnosis judging module judges that tongue diagnosis needs to be combined, the final syndrome and/or the syndrome are obtained by integrating the symptoms and tongue picture information of the patient based on the syndrome and the deep learning network model; can greatly improve the effect of inquiry and provide powerful support for syndrome differentiation and typing.

Description

Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning
Technical Field
The disclosure relates to the technical field of traditional Chinese medicine inquiry system, in particular to a traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The inquiry of traditional Chinese medicine is the core of the four diagnostic methods of traditional Chinese medicine, such as inspection, smelling, inquiry and center-cutting, and doctors of traditional Chinese medicine need to inquire about some problems of patients in a targeted manner according to symptoms described by the patients and by combining with the knowledge of traditional Chinese medicine and the experience of going to the doctor, so as to achieve the purpose of rapidly judging the syndrome and differentiating the types of syndromes. At present, an automatic inquiry system is mainly based on theories of traditional Chinese medicine constitution identification questionnaires, traditional Chinese medicine shits songs and the like, asks a large number of similar questions for different patients, and comprehensively collects the health conditions of all aspects of users.
On the other hand, the tongue diagnosis is an important component of inspection, auscultation, inquiry and palpation of TCM, and doctors in TCM need to comprehensively observe and hear the results of the four diagnostic methods in the face of patients and deduce the syndrome and syndrome based on their own medical experiences. Therefore, how to combine the tongue diagnosis result and the result based on the symptom inquiry is an important issue that each doctor of TCM must face. Most of the traditional Chinese medical artificial intelligence auxiliary diagnosis and treatment systems at the present stage only treat the tongue picture as a symptom, or give different weights to the tongue picture and the symptom mathematically in the process of syndrome differentiation and typing, and do not deeply explore the relationship between the tongue picture and the symptom under different conditions by combining the traditional Chinese medical theory and experience with a mathematical model.
The inventor of the present disclosure finds that the existing traditional Chinese medicine inquiry technology has the following defects: (1) In the current stage, the traditional Chinese medicine automatic inquiry system is carried out based on theories of traditional Chinese medicine constitution identification questionnaires, traditional Chinese medicine shits and the like, the problem homogenization phenomenon is serious, namely, the questions asked for by different patients are basically consistent (occasionally slightly different according to sex and age stages), the traditional Chinese medicine inquiry can not be carried out in a targeted manner according to the symptoms of the patients, and the use experience of the patients is poor. (2) At present, the traditional Chinese medicine automatic inquiry system on the market asks a large number of questions to patients, and because too many questions are asked and the questions are not targeted, the patients are easily misled by the questions in the answering process to provide interference information, so that the inquiry effect is influenced, and strong evidence cannot be provided for syndrome differentiation and typing. In addition, the current automatic traditional Chinese medicine inquiry system does not combine the symptoms of the patient with the experience of the famous traditional Chinese medicine family to provide thinking and directions for inquiry, thereby influencing the inquiry effect. (3) The relationship between tongue manifestation and symptoms is very complicated in the field of TCM, and if the tongue manifestation is treated as a symptom, the function of tongue diagnosis in the process of TCM diagnosis and treatment cannot be highlighted, thereby affecting the result of TCM diagnosis. (4) The traditional Chinese medicine artificial intelligence auxiliary diagnosis and treatment system does not use artificial intelligence technology to carry out mathematical modeling on the relationship of tongue manifestation and symptoms under different conditions, and does not provide different combination modes by combining the experience of famous old Chinese medicine, thereby influencing the diagnosis result of the Chinese medicine.
Disclosure of Invention
In order to solve the defects of the prior art, the invention provides a traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome elements and deep learning, which aims to analyze the symptoms of a patient, deeply fuses the deep learning technology with the experience of famous Chinese medicine to deduce the possible syndrome of the patient, and then carries out accurate inquiry through the traditional Chinese medicine syndrome element theory to confirm the deduction of a mathematical model.
In order to achieve the purpose, the following technical scheme is adopted in the disclosure:
the first aspect of the disclosure provides a traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome and deep learning.
A traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome and deep learning comprises:
a symptom analysis module configured to: analyzing symptom information input by a patient according to a preset knowledge graph based on the syndrome of the symptom;
an interrogation determination module configured to: judging whether an inquiry is required or not based on the result of symptom analysis;
an interrogation module configured to: if the judgment result is that the inquiry is needed, the traditional Chinese medicine inquiry is carried out by utilizing the syndrome and the deep learning network model, and the syndrome and/or the syndrome of the patient are/is obtained according to the answer of the patient;
a tongue diagnosis determination module configured to: judging whether tongue diagnosis is combined, if so, entering the result of the inquiry module into a combined tongue diagnosis module to obtain the final syndrome and/or syndrome element; otherwise, the result obtained by the inquiry module is used as the final syndrome and/or syndrome element;
a binding tongue module configured to: when the inquiry judging module judges that the inquiry is not needed or the tongue diagnosis judging module judges that the tongue diagnosis needs to be combined, the final syndrome and/or the syndrome are obtained by integrating the symptoms and the tongue picture information of the patient based on the syndrome and the deep learning network model.
As some possible implementations, the knowledge-graph includes:
a first knowledge-graph configured to: the syndrome corresponding to each symptom;
a second knowledge-graph configured to: categories corresponding to each symptom;
a third knowledge-graph configured to: aiming at typical symptoms corresponding to a syndrome in each symptom category;
a fourth knowledge-graph configured to: the corresponding syndrome element of each syndrome;
a fifth knowledge-graph configured to: including composite syndrome and ranking;
a sixth knowledge-graph configured to: the corresponding syndrome of each tongue.
By way of further limitation, the symptom analysis module comprises:
transforming symptom descriptions input by the patient into various symptoms stored in a first knowledge graph to obtain a first syndrome set;
aiming at each symptom in the first syndrome set, finding out the disease syndrome and the morbid syndrome corresponding to each symptom in the first knowledge graph, and sequencing all the appeared disease syndromes and morbid syndromes according to the frequency and the sequence of appearance in the symptom;
selecting a plurality of syndrome elements which are ranked most front from the disease syndrome elements and the disease location syndrome elements respectively to form a second syndrome element set;
and (3) if all disease symptoms and disease location symptoms corresponding to all symptoms in the first symptom set do not appear in the second symptom set, combining the symptoms into a third symptom set, and sorting according to the frequency and the sequence of the appearance in the symptoms.
As a further limitation, in the inquiry judging module, it is judged whether the evidences in the second evidences set can explain the chief complaints, and the evidences included in the second evidences set can explain a certain symptom, specifically:
for the symptom, if the intersection of the disease syndrome set and the disease location syndrome set corresponding to the symptom in the first knowledge graph and the intersection of the disease syndrome set and the second syndrome set are not empty, the syndrome contained in the second syndrome set is called to explain the symptom, and the inquiry is not performed at this moment, and the tongue picture judgment module is entered; otherwise, the system enters an inquiry module for inquiry.
As a further limitation, the interrogation module comprises:
starting with the first disease viseme in the third viseme set;
for any disease symptom element in the third syndrome element set, finding a symptom corresponding to the disease symptom element in the first syndrome element set, finding a category corresponding to the symptom element in the second knowledge graph, finding a typical syndrome element set of the syndrome element in the category in the third knowledge graph, and inquiring whether a certain symptom in the typical syndrome element set appears;
if the answer of the user is yes, selecting the disease condition element, bringing the disease condition element into the second condition element set, if the condition element contained in the new second condition element set can explain the chief complaint, stopping the inquiry, outputting the second condition element set, and entering the tongue picture judgment module;
if the evidence element contained in the new second evidence element set cannot explain the chief complaint or the answer of the user is no, the step is executed again for the next evidence element in the third evidence element set;
if the steps are carried out for all the syndrome elements in the third syndrome element set, the syndrome elements contained in the second syndrome element set can not explain the chief complaints, and the inquiry and diagnosis are carried out by combining the experience of the famous old Chinese medicine.
As a further limitation, the inquiry module, combined with the inquiry of the experience of the old traditional Chinese medicine, comprises:
inputting the first evidence set into a deep learning network model, recommending a plurality of candidate syndromes through the deep learning network model, and carrying out the following operations on any candidate syndrome from the first candidate syndrome:
finding out a disease property syndrome set corresponding to the current syndrome in the fourth knowledge graph, stopping inquiry if the disease property syndromes in the disease property syndrome set are in the second syndrome set, outputting the current syndrome, and entering a tongue picture judgment module; otherwise, composing the disease visemes in the disease viseme set but not in the second viseme set into a fourth viseme set;
starting from the first disease property element in the fourth element set, inquiring each disease property element whether a user has a certain symptom in the typical element set, if so, outputting the current syndrome and entering a tongue picture judgment module; if the user answers no to a certain symptom element, the first two steps of operation are carried out for the next syndrome;
if any one of the candidate syndromes cannot be output through the three steps, outputting a second syndrome set and entering a tongue picture judgment module.
As a further limitation, if the results of the inquiry judging module and the inquiry module are certain syndromes, the obtained results are directly used as final syndromes; otherwise, substituting the second evidence collection output by the inquiry judging module and the inquiry module into the combined tongue diagnosis module;
the combined tongue diagnosis module comprises:
converting the tongue picture description of the patient into various tongue pictures stored in a sixth knowledge graph of the traditional Chinese medicine to obtain a tongue picture set;
the second syndrome set obtained by the inquiry module comprises a first disease syndrome set and a first disease location syndrome set;
aiming at each tongue picture in the tongue picture set, finding out the disease property syndrome and disease location syndrome corresponding to each tongue picture in the sixth knowledge graph to obtain a tongue picture syndrome set and obtain a second disease property syndrome set and a second disease location syndrome set;
the intersection of the second syndrome element set and the tongue syndrome element set obtained by the inquiry module is a first syndrome element intersection, the intersection of the first disease syndrome element set and the second disease syndrome element set is a second syndrome element intersection, and the intersection of the first disease syndrome element set and the second disease syndrome element set is a third syndrome element intersection.
As a further limitation, if the intersection of the second credentials is empty, the method includes:
obtaining a plurality of candidate syndromes through a deep learning network model to obtain a candidate syndrome set, and starting from the first candidate syndrome, entering the next step for any candidate syndrome;
finding out a disease character set corresponding to the candidate syndrome in the fourth knowledge graph, and entering the next step;
if the intersection of the disease property syndrome set corresponding to the current candidate syndrome and the first disease property syndrome set is not a null set, outputting the current candidate syndrome; otherwise, the operation of the previous step is carried out aiming at the next candidate syndrome;
and if the disease property syndrome set corresponding to all the candidate syndromes does not intersect with the first disease property syndrome set, outputting a second syndrome set by taking the syndrome derived from the symptoms as the main.
As a further limitation, if the second viseme intersection contains only one disease viseme, including:
deducing a plurality of candidate syndromes through a deep learning network model to obtain a candidate syndrome set, defining a null set K from a first candidate syndrome, and entering the next step for any candidate syndrome;
if the number of syndromes in K is less than the preset number, finding out a disease character set corresponding to the current candidate syndromes in the fourth knowledge graph, and entering the next step;
if the disease syndrome element set and the second syndrome element intersection have intersection, adding the current candidate syndrome into the set K, and then performing the operation in the previous step aiming at the next candidate syndrome; if the disease syndrome element set and the second syndrome element intersection do not have intersection, the operation in the previous step is carried out aiming at the next candidate syndrome; if the current candidate syndrome is the last syndrome in the candidate syndrome set, entering the next step;
if the number of the syndromes in K reaches a preset value and K is an empty set, outputting a first syndrome intersection, if K is not an empty set, finding a disease property syndrome set corresponding to the current syndrome in K in a fourth knowledge graph, and starting from the first syndrome in K, and entering the next step for each syndrome;
if the intersection of the disease viseme set corresponding to the current candidate viseme in the K is equal to the second viseme, outputting the current candidate viseme in the K, and otherwise, entering the next step;
for each disease viscus which is contained in the disease viscus set corresponding to the current candidate viscus in the K and is not in the second viscus set, finding typical symptoms corresponding to the viscus in a third knowledge graph, inquiring whether a user has one of the typical symptoms, and if the typical symptoms of each disease viscus are answered by the user, outputting the current viscus; otherwise, if the syndrome is not the last syndrome in K, the operation in the previous step is performed on the next syndrome in K, and if the syndrome is the last syndrome in K, the first syndrome intersection is output.
As a further limitation, if the second viseme intersection contains at least two disease visemes:
according to the ranking of the composite syndromes in the fifth knowledge graph, starting from the first composite syndrome, obtaining a disease property syndrome set corresponding to the current composite syndrome for any one composite syndrome, and entering the next step;
if the second syndrome intersection is equal to the set of disease-related syndromes corresponding to the current compound syndrome, outputting the current compound syndrome; if the second syndrome intersection is not equal to the disease property syndrome set corresponding to the composite syndrome, and the second syndrome intersection is a subset of the disease property syndrome set corresponding to the current composite syndrome, performing the next step;
for each disease viscidity element contained in the disease viscidity element set corresponding to the compound viscidity but not in the second viscidity element set, finding typical symptoms corresponding to the viscidity elements in a third knowledge map, inquiring whether a user has one of the typical symptoms or not, and entering the next step;
if the user answers yes to the typical symptoms of each disease viseme, outputting the current compound syndrome; otherwise, the operation in the previous step is carried out according to the next composite syndrome;
if the second syndrome intersection is not the subset of the disease syndrome set corresponding to the composite syndrome, performing the operation in the second step on the next composite syndrome;
and if the composite syndrome cannot be output, outputting the intersection of the first syndrome elements.
A second aspect of the present disclosure provides a medium having a program stored thereon, the program, when executed by a processor, implementing the steps of:
the symptom analysis module analyzes symptom information input by a patient according to a preset knowledge graph based on the syndrome of the symptom;
the inquiry judging module judges whether inquiry is needed or not based on the result of symptom analysis;
if the judgment result is that the inquiry is needed, the inquiry module utilizes the syndrome and the deep learning network model to carry out traditional Chinese medicine inquiry, and obtains the syndrome and/or the syndrome of the patient according to the answer of the patient;
the tongue diagnosis judging module judges whether the tongue diagnosis is combined, if so, the result of the inquiry module enters the combined tongue diagnosis module to obtain the final syndrome and/or syndrome element; otherwise, the result obtained by the inquiry module is used as the final syndrome and/or syndrome element;
when the inquiry judging module judges that inquiry is not needed or the tongue diagnosis judging module judges that tongue diagnosis needs to be combined, the tongue diagnosis module is combined to synthesize the symptoms and tongue picture information of the patient based on the syndrome and the deep learning network model to obtain the final syndrome and/or the syndrome.
A third aspect of the present disclosure provides an electronic device, including a memory, a processor, and a program stored on the memory and executable on the processor, where the processor executes the program to implement the following steps:
the symptom analysis module analyzes symptom information input by a patient according to a preset knowledge graph based on the syndrome of the symptom;
the inquiry judging module judges whether inquiry is needed or not based on the result of symptom analysis;
if the judgment result is that the inquiry is needed, the inquiry module uses the syndrome and the deep learning network model to carry out traditional Chinese medicine inquiry, and obtains the syndrome and/or the syndrome of the patient according to the answer of the patient;
the tongue diagnosis judging module judges whether the tongue diagnosis is combined, if so, the result of the inquiry module enters the combined tongue diagnosis module to obtain the final syndrome and/or syndrome element; otherwise, the result obtained by the inquiry module is used as the final syndrome and/or syndrome element;
when the inquiry judging module judges that the inquiry is not needed or the tongue diagnosis judging module judges that the tongue diagnosis needs to be combined, the tongue diagnosis module is combined to synthesize the symptoms and tongue picture information of the patient based on the syndrome and the deep learning network model to obtain the final syndrome and/or the syndrome.
Compared with the prior art, the beneficial effect of this disclosure is:
1. the system, the medium or the electronic equipment automatically complete the inquiry based on the symptoms and the tongue picture information of the patient, combine the tongue picture information in different modes under different conditions to infer the syndrome and/or the syndrome of the patient, analyze the disease nature syndrome of the patient according to the symptoms of the patient, perform targeted symptom inquiry, and enable the patient to experience the intelligent characteristic of the system deeply so as to obtain good use experience.
2. Compared with the traditional Chinese medicine automatic inquiry system, the medium or the electronic equipment can greatly improve the inquiry effect and provide strong support for dialectical analysis.
3. The system, the medium or the electronic equipment comprehensively analyze the symptoms and the tongue picture of the patient, process the symptoms and the tongue picture in different ways according to different conditions, deeply fuse the traditional Chinese medicine knowledge and the mathematical model and greatly improve the effect of comprehensively analyzing the symptoms and the tongue picture of the patient.
4. The system, the medium or the electronic equipment disclosed by the disclosure combines the tongue picture and the symptom of the patient with the experience of the famous Chinese medicine through a deep learning model, finds out possible syndromes for the patient, and then performs automatic supplementary inquiry around the syndrome of the possible syndromes, thereby greatly improving the accuracy of Chinese medicine diagnosis.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure and are not to limit the disclosure.
Fig. 1 is a schematic structural diagram of a traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome and deep learning provided in embodiment 1 of the present disclosure.
Fig. 2 is a schematic diagram of a knowledge-graph structure provided in example 1 of the present disclosure.
Fig. 3 is a map structure of "symptom → symptom category" in the knowledge map T provided in example 1 of the present disclosure.
Fig. 4 is a schematic workflow diagram of the intelligent traditional Chinese medicine tongue diagnosis integrated system based on syndrome and deep learning provided in embodiment 1 of the present disclosure.
Fig. 5 is a schematic workflow diagram of a model M provided in embodiment 1 of the present disclosure.
Fig. 6 is a schematic workflow diagram of a model D provided in embodiment 1 of the present disclosure.
Detailed Description
The present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments in the present disclosure may be combined with each other without conflict.
Example 1:
as shown in fig. 1, embodiment 1 of the present disclosure provides a traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome and deep learning, which includes the following processes:
(1) In the "symptom analysis module," the patient-entered symptom information is analyzed based on the syndrome of the symptom.
(2) Based on the result of the symptom analysis, the "module for determining whether to perform an inquiry" determines whether to perform an inquiry.
(3) If the result is that the inquiry is needed, based on the symptoms of the patient, the traditional Chinese medicine inquiry is automatically carried out in an inquiry module by utilizing the syndrome and the deep learning technology, and the syndrome and/or the syndrome of the patient are/is output according to the answer of the patient. And then, judging whether to enter a combined tongue diagnosis module or not based on a judgment result in the inquiry judgment module or not, if so, entering the combined tongue diagnosis module, and otherwise, directly outputting a final result in an output module.
(4) If the judgment result in the inquiry judging module is that the inquiry is not needed, the method enters a combined tongue diagnosis module.
(5) The 'combined tongue diagnosis module' comprehensively analyzes the symptoms and tongue picture information of the patient based on the syndrome and the deep learning technology, and finally outputs the syndrome and/or the syndrome in the 'output module'.
Specifically, the method comprises the following steps:
1. description of the concept
(1) Evidence element: the minimal unit describing body lesions in the theory of traditional Chinese medicine is divided into two types, namely pathotype element and pathotype element, wherein the pathotype element refers to lesions (such as yin deficiency and yang deficiency) appearing in the body, the pathotype element refers to the positions (such as kidney) where the lesions appear, and the pathotype element mentioned in the application comprises both the pathotype element and the pathotype element.
(2) The syndrome: in traditional Chinese medicine, the diagnosis is equivalent to that of western medicine, and consists of one or more disease symptoms combined (or not combined) with one (or more) disease symptom elements, for example, the syndrome of deficiency of both yin and yang of kidney consists of disease symptoms of yin deficiency, yang deficiency and disease symptom elements of kidney, and the syndrome of yin deficiency consists of disease symptoms of yin deficiency.
(3) The compound syndromes: the syndrome composed of at least two kinds of pathological factors is called a complex syndrome.
(4) Typical symptoms of the syndrome: each symptom corresponds to one or more (disease property or disease position) elements, and if a certain symptom only corresponds to one (disease property or disease position) element, the symptom is called as a typical symptom of the element.
(5) Symptom categories: each symptom belongs to a certain category, for example, the symptom "right upper abdominal pain" belongs to the category "abdominal" and the symptom "insomnia" belongs to the category "sleep".
(6) Tongue manifestation: including the tongue proper, coating, and tongue shape, e.g., purple tongue, thin and white coating, and tooth mark on tongue.
2. Carrying out the step
(1) A traditional Chinese medicine knowledge graph is created and stored in a graph database neo4 j. A total of six knowledge-maps are created, each including the following category correspondences.
Knowledge graph P: symptom → syndrome element, i.e. the syndrome element corresponding to a certain symptom.
Knowledge graph T: symptom → symptom category, i.e. the category corresponding to a certain symptom.
Knowledge graph S: syndrome → symptom category → typical symptom, i.e. for a syndrome, it corresponds to the typical symptom in each symptom category.
Knowledge graph W: syndrome → syndrome element, i.e. the syndrome element corresponding to a certain syndrome.
Knowledge graph F: the ranking of the composite syndrome → the ranking of the composite syndrome here means that all the composite syndromes are ranked according to the number of times of occurrence of big data statistics, and the more common the composite syndromes are ranked the earlier.
Knowledge graph N: tongue → Zhen Su, i.e. the corresponding Zhen Su of a certain tongue.
The structure of the knowledge-graph is shown in fig. 2, wherein node types represent different classes in the graph and relationships represent relationships between the different classes.
For example, the atlas structure of "symptom → symptom category" in the knowledge atlas T is shown in fig. 3.
In the following, in steps 2-9, a system implementation flow is described, and a flow chart is shown in fig. 4 (wherein the numbers represent the sequence numbers of the steps):
in step 2, the natural language text input by the patient is converted into Chinese medicine symptom terms through the knowledge graph P, and the corresponding syndrome elements of the symptoms are obtained based on the knowledge graph P, and then the step 3 is carried out.
In step 3, judging whether the evidence element obtained in step 2 can explain the chief complaint, if not, entering step 4 to perform inquiry, otherwise, entering step 6.
In step 4, performing inquiry diagnosis based on the knowledge map T, S, judging whether further inquiry diagnosis of the experience of the famous old Chinese medicine is needed in step 4.3, and if so, entering step 5; otherwise, go to step 6
In step 5, based on the knowledge map W and the deep learning model M (the model M is described in step 10), the inquiry based on the experience of the famous old Chinese medicine is carried out, and the procedure goes to step 6.
In step 6, judging whether tongue picture information needs to be combined or not, if so, entering step 7, otherwise, entering step 9 and outputting a result.
In step 7, the tongue image input by the patient is analyzed based on the knowledge-graph N, and the process proceeds to step 8. In step 8, the symptoms and tongue manifestation of the patient are comprehensively analyzed based on the deep learning model M and the knowledge map F, and the results are output in step 9.
The detailed steps are as follows:
step 2: analyzing patient-entered symptoms
Step 2.1: first, the symptom description inputted by the patient is converted into various symptoms stored in the knowledge graph P of traditional Chinese medicine. For this purpose, the set of all symptoms in the knowledge graph P is converted into a symptom dictionary; the symptom description input by the patient is called a meta-sentence, and the meta-sentence is called a compound sentence after punctuation is removed. Next, the following operations are performed:
step 2.1.1: the segmented sentence is segmented using the resultant segmentation in combination with the symptom dictionary in 2.1, and the segmentation result is called an initial segmentation set.
Step 2.1.2: inquiring each word in the initial word segmentation set in a symptom dictionary, and if the word can be found, bringing the words into an output symptom set; words that cannot be found are collected and are called as a word set to be processed.
Step 2.1.3: for each word i in the word set to be processed, finding out the punctuation where the word i is in the meta sentence, converting the punctuation into pinyin (only pinyin letters without tones), finding out the symptom with the highest score in the pinyin set containing all symptoms in the symptom dictionary by utilizing a search matching technology Lucene in combination with a BM25 algorithm, and taking the symptom as the recognition result of i to be included in the output symptom set.
Step 2.1.4: the set of symptoms obtained in 2.1.1-2.1.3 was recorded as A and step 2.2 was entered.
Step 2.2: for each symptom in A, finding out the disease property element and disease location element corresponding to each symptom in the knowledge graph P, and sequencing all the appeared disease property elements and disease location elements according to the frequency and sequence of appearance in the symptom (namely, arranging the elements with more appearance frequencies in front, and sequencing the elements with the same appearance frequency according to the sequence of appearance).
Step 2.3: for the disease syndrome and the disease location syndrome, several (for example, 3) syndromes with the highest ranking are selected, and the set of these syndromes is denoted as C.
Step 2.4: if all disease evidences and disease location evidences corresponding to all symptoms in the symptom set A do not appear in the set C, the evidences are combined into a set B, and the evidences in the set B are sorted according to the sorting method of 2.2.
Step 2.5: step 3 is entered.
And step 3: judging whether to perform inquiry
Refer to the definition of set C in step 2.4.
Step 3.1: judging conditions: the syndrome contained in set C can explain the chief complaint (e.g., the first 5 symptoms entered by the patient are defined as the chief complaint).
Step 3.2: concept interpretation: the definition that the syndrome contained in the set C can explain a certain symptom is that, regarding the symptom, the corresponding disease syndrome set P _ x and disease syndrome set P _ w in the knowledge graph P are considered, and if the intersection of the set C and the P _ x and the P _ w are not empty, the syndrome contained in the set C is called to explain the symptom. If the syndrome element contained in set C can explain all symptoms in the chief complaint, the syndrome element contained in set C is said to explain the chief complaint.
Step 3.3: and (3) judging a rule: if the judgment condition in 3.1 is met, the inquiry is not carried out, the set C is output, and the step 6 is entered; otherwise, go to step 4 to perform inquiry.
And 4, step 4: interrogation for syndrome set C
Refer to the definition of set A, B, C in step 2.3.
Step 4.1: starting with the first disease viseme B _1 in B, for each disease viseme B _ i in B, step 4.2 is entered.
Step 4.2: for the evidence element B _ i in the set B, finding the symptom a _ i corresponding to the B _ i in the set A, finding the category T _ i corresponding to the a _ i in the knowledge graph T, finding the typical symptom set S _ i of the B _ i in the category T _ i in the knowledge graph S, and inquiring whether a certain symptom in the set S _ i appears.
Step 4.3: in step 4.2, if the user answers yes, selecting a disease viseme b _ i, bringing the disease viseme b _ i into the set C, if the new set C meets the condition 3.1, stopping the inquiry, outputting the set C, and entering step 6, otherwise, entering step 4.4; if in step 4.2 the user answers no, step 4.4 is entered.
Step 4.4: for the next credential B _ (i + 1) in set B, step 4.2 is entered.
Step 4.5: if steps 4.2-4.4 are performed for all the certificates in B, the condition 3.1 still cannot be met, and the procedure goes to step 5.
And 5: inquiry combined with famous and old traditional Chinese medicine experiences
Entering the step, the inquiry made according to the traditional Chinese medicine knowledge in the knowledge map is shown to be incapable of effectively explaining the symptoms of the patient (namely, the condition 3.1 cannot be met), and the step combines the famous and old traditional Chinese medicine experiences (embodied in the form of the model M described in the step 10) to provide an idea for inquiry. Refer to the definition of set A, C, respectively, in step 2.4. Inputting the symptom set A into a model M (see step 10), recommending n (n =5 can be assumed) candidate syndromes through the model M (the candidate syndrome set is R, n syndromes in R are ranked according to the recommended sequence and are recorded as R _1 and R _2, namely the ith syndrome in the set is recorded as R _ i), and starting from the first candidate syndrome R _1, performing the following operation on each R _ i:
step 5.1: finding out a disease syndrome set E _ i (wherein the jth syndrome is marked as E _ (i, j)) corresponding to the syndrome R _ i in the knowledge graph W, if the disease syndromes in the E _ i are all in the set C, stopping the inquiry, outputting the syndrome R _ i, and entering the step 6; otherwise, the set composed of disease symptoms in the set E _ i but not in the set C is recorded as F _ i (wherein the jth disease symptom is recorded as F _ (i, j)), and step 5.2 is entered.
Step 5.2: starting from f _ (i, 1), the operation in step 4.2 is performed for each f _ (i, j), and step 5.3 is entered.
Step 5.3: if the answer of the user is yes for each f _ (i, j), outputting a syndrome R _ i, and entering the step 6; if the user answers no for one f _ (i, j), the operations in steps 5.1-5.2 are performed for the syndrome R _ (i + 1).
Step 5.4: and if any syndrome in the R cannot be output after the steps of 5.1-5.3, outputting a syndrome set C, and entering the step 6.
Step 6: determine whether to combine tongue information
If the result output in the step 1-5 is a certain syndrome, directly entering the step 9; otherwise, step 7 is entered, and the certificate collection C output in steps 1-5 is substituted into step 7.
And 7: analyzing tongue picture input by patient
Step 7.1: and (3) converting the tongue picture description of the patient into various tongue pictures stored in the knowledge graph N of the traditional Chinese medicine according to the method in the step 2.1, and recording a set consisting of the converted tongue pictures as T _ o.
Step 7.2: referring to the syndrome set C obtained in step 6 based on symptoms (and inquiry), the set of disease syndrome components contained in C is denoted as C _ x, and the set of disease site syndrome components contained in C is denoted as C _ w.
Step 7.3: for each tongue picture in the T _ o, finding out the disease syndrome and the morbid syndrome corresponding to each tongue picture in the knowledge graph N, and recording a set formed by the syndrome as Y, a set formed by the disease syndrome contained in Y as Y _ x, and a set formed by the morbid syndrome contained in Y as Y _ w.
Step 7.4: and (5) recording the intersection of C and Y as Z, the intersection of C _ x and Y _ x as Z _ x, and the intersection of C _ w and Y _ w as Z _ w, and entering the step 8.
And 8: comprehensive analysis of symptoms and tongue manifestation
Step 8.1: if Z _ x is an empty set
Step 8.1.1: n candidate syndromes (n =5 can be assumed) are deduced through the model M (see step 10), and a set R is formed (n syndromes in R are ranked according to the recommended precedence order and are denoted as R _1, R _2. Starting from R _1, for each R _ i, go to step 8.1.2.
Step 8.1.2: and finding a disease syndrome set E _ i corresponding to the R _ i in the knowledge graph W, and entering the step 8.1.3.
Step 8.1.3: if the intersection of the E _ i and the C _ x is not an empty set, outputting a syndrome R _ i, and entering the step 9; if the intersection of E _ i and C _ x is empty, then the operation in step 8.1.2 is performed for R _ (i + 1).
Step 8.1.4: if all E _ i and C _ x have no intersection, the syndrome set C is output by taking the syndrome derived from the symptom as the main part, and the step 9 is entered.
Step 8.2: if Z _ x contains only one disease syndrome
Step 8.2.1: n candidate syndromes are deduced through the model M to form a set R (the n syndromes in the R are ranked according to the recommended sequence and are recorded as R _1 and R _2, namely the ith syndrome in the set is recorded as R _ i). Defining K as an empty set, starting from R _1, for each R _ i, go to step 8.2.2.
Step 8.2.2: if the number of syndromes in K reaches t (assuming t = 5), go to step 8.2.4. If the number of syndromes in K is less than t, finding out a disease syndrome set E _ i corresponding to R _ i in the knowledge graph W, and entering the step 8.2.3.
Step 8.2.3: if E _ i intersects Z _ x, then R _ i is added to the set K, and the operation in step 8.2.2 is performed for R _ (i + 1). If E _ i does not intersect Z _ x, the operation in step 8.2.2 is performed for R _ (i + 1), and if R _ i is already the last syndrome in R, step 8.2.4 is entered.
Step 8.2.4: if K is empty, outputting a certificate set Z and entering step 9. If the K is not an empty set, the ith syndrome in the K is recorded as K _ i, a disease syndrome set W _ i corresponding to the K _ i is found in the knowledge graph W, and the step 8.2.5 is carried out on each W _ i from W _ 1.
Step 8.2.5: and if W _ i is equal to Z _ x, outputting the syndrome K _ i. If W _ i is not equal to Z _ x, proceed to 8.2.6.
Step 8.2.6: for each disease syndrome contained in W _ i but not in Z _ x, the typical symptoms (of all symptom categories) corresponding to these syndromes are found in the knowledge graph S, and the user is asked whether or not to have one of these typical symptoms. If the user answers yes to the typical symptoms of each disease syndrome, outputting a syndrome K _ i; otherwise, if K _ i is not the last syndrome in K, the operation in step 8.2.5 is performed for W _ (i + 1), and if K _ i is the last syndrome in K, the syndrome set Z is output, and the process proceeds to step 9.
Step 8.3: if Z _ x contains at least two pathological markers
Let the disease ID in Z _ x be Z _1, Z _2.
Step 8.3.1: according to the ranking of the composite syndromes in the knowledge graph F (the composite syndrome with the ranking of i in F is marked as F _ i, and the disease property syndrome set corresponding to the composite syndrome is marked as D _ i)), starting from the first composite syndrome F _1, the step proceeds to step 8.3.2 for each F _ i.
Step 8.3.2:
1) If Z _ x is equal to D _ i, outputting syndrome F _ i.
2) If Z _ x is not equal to D _ i and Z _ x is a subset of D _ i, go to step 8.3.3.
3) If Z _ x is not a subset of D _ i, proceed to step 8.3.5.
Step 8.3.3: for each disease syndrome contained in D _ i but not in Z _ x, the typical symptoms (of all symptom categories) corresponding to these syndromes are found in the knowledge graph S, the user is asked whether any one of these typical symptoms appears, and the process proceeds to step 8.3.4.
Step 8.3.4: if the user answers yes for each typical symptom of the disease syndrome, outputting a syndrome F _ i; otherwise, the operation in step 8.3.3 is performed again for F _ (i + 1).
Step 8.3.5: the operation in step 8.3.2 is performed for the composite syndrome F _ (i + 1).
Step 8.3.6: if the syndrome F _ i cannot be output through 8.3.1-8.3.5, outputting the syndrome set Z and entering step 9.
Step 8.4: based on the results output from 8.1-8.3, the syndrome and/or syndrome set is output.
And step 9: and (3) outputting the syndromes and/or syndrome sets obtained in the steps 1-8.
Step 10: model M
Step 10.1: introduction to function
The model deduces possible syndromes of the patient based on various types of information of the patient. Specifically, the input of the model is various types of information of the patient, the output is n possible syndromes, and the n syndromes are ranked according to the probability. In steps 5 and 8, two models are mentioned, the basic structure and output of the two models are the same, but the input information is different, and in the two models, the input information comprises the basic information, disease description, syndrome and western medicine diagnosis of the patient, but the disease description is different, and the specific details are as follows:
in step 5, the condition description input to the model M includes a symptom description of the patient;
in step 8, the disease description input to the model M includes a description of the patient's symptoms and a description of the tongue.
In this example we will describe this model uniformly and in the 10.2 explanation of the model we will describe this concept uniformly using the disease, which in the two models mentioned in steps 5, 8 represents the two different meanings mentioned above.
Note: in step 10.2, in (3), regarding the ordering of disease descriptions, if the disease description includes both symptom descriptions and tongue picture descriptions, the first three symptoms of the disease description are defined as follows, the first symptom description is ranked first in the disease description, the second symptom description is ranked second in the disease description, and the first tongue picture description is ranked third in the disease description.
Step 10.2: model construction
Step (1): and (3) constructing a famous old traditional Chinese medicine experience information base H, wherein the information base H contains basic information, disease description, syndrome, western medicine diagnosis and the like of different patients, and the information base H is used for constructing a model D, G mentioned in the following steps (2) and (3).
Step (2): recording all information of the patient as A, inputting A into a deep learning model D, deducing m (m is not less than n) candidate syndromes, forming a set D _ A by the m candidates, and respectively assigning scores with the values of m, m-1,.. 2 and 1 to the m candidate syndromes according to the sequence when the model is output.
And (3): the first three descriptions in the disease description set A are respectively input into a statistical model G to respectively deduce m candidate syndromes, the three groups of syndromes respectively form sets G _1, G _2 and G _3, and the m candidate syndromes in each set are respectively given scores with the numerical values of m, m-1,. 2 and 1 according to the sequence when the models are output.
And (4): the scores of the syndromes appearing in D _ A, G _1, G _2 and G _3 are respectively given the weights of 0.4, 0.3, 0.2 and 0.1, the weighted sum of the scores of all syndromes is calculated, the n syndromes with the highest scores are found, and the n syndromes are output.
The workflow of model M is shown in fig. 5.
The specific implementation steps of the step (1) are as follows:
a step (101): for each syndrome in the knowledge graph W, k (e.g., k = 100) "different patients" having the syndrome are found in the famous-old-chinese-medical-case library (see definition step (102)), and the sex, age, disease description, syndrome, and name of western-medical-disease diagnosis (if there is no name of western-medical-disease diagnosis in the case library, it is recorded as "no western-medical-disease diagnosis") of the patients are stored in the information library H. The information base H has a storage structure of syndrome name → patient number (k patients per syndrome name) → sex, age, description of condition, syndrome, and name of western medicine disease diagnosis of the patient with a certain number.
A step (102): the definition of "different patients" is that the similarity between the disease description texts of each patient, calculated using the Levenshtein distance, is less than a certain threshold, such as 0.5.
The specific implementation steps of the step (2) are as follows:
step (201): and constructing an information vector corresponding to each patient in the information base H.
For patient p, the information mainly includes two types, one is structured information (including sex, age, syndrome, and name of western medicine disease diagnosis of patient), and the other is unstructured text information (including description of patient's condition).
(i) For structured information, we convert the structured information into vector information (called category vector) according to 0-1 vector corresponding to the category, i.e. a certain vector can be used to represent the structured information, for example, as follows:
example 1, the gender of the patient is two types (male and female), the male category vector is (1,0) and the female category vector is (0,1), then the gender information of the patient is converted from "male or female" to a vector (1,0) or (0,1) ".
Example 2, assuming that the information base H has 3000 western medicine disease diagnosis names in total, the category vectors corresponding to the disease diagnosis names of the patients have 3000 dimensions, the dimension values corresponding to the disease diagnoses of the patients are 1, and the remaining dimension values are 0.
For patient p, all structured information is converted into a category vector, and all category vectors of the patient are combined into one vector, which is denoted as L _ p.
(ii) For the unstructured text information, vectorization representation of the illness description text is completed, namely, words in the illness description are represented by word vectors by word segmentation technologies (such as Chinese segmentation), word vector technologies (such as word2vec, BERT and the like).
Step (202): the procedure for constructing model D is as follows.
(i) Establishing a training set, a verification set and a test set
For k patients corresponding to each syndrome in the patient condition information base H, the information vectors corresponding to 60%, 20% and 20% of the patients are respectively classified into a training set, a verification set and a test set.
(ii) Structure of deep learning model D
Inputting a vector L _ p of structured information of a patient p into a multilayer perceptron model, and recording the output vector as M _ p;
inputting the (possibly a plurality of) unstructured text information word vectors of the patient p into a bidirectional long-time memory recurrent neural network model, and obtaining a vector C _ p capable of expressing the disease description of the patient p by passing output results of all output word vectors through a maximum pool technology;
defining M _ p and C _ p in combination as a vector V _ p = [ M _ p, C _ p ];
mapping the vector V _ p to an exponential vector p representing the similarity relation between the information of the patient p and all possible syndromes by using a linear regression method, namely, rho = c _ 1V _p _c _2, wherein the dimension of rho is the number of all possible syndromes in the knowledge graph W, the ith element rho (i) in the vector p represents the relation between the information of the patient p and the ith syndrome in the knowledge graph W, wherein c _1 and c _2 are parameter tensors in the linear regression relation, and the training of the following step (iii) is needed to determine a specific numerical value;
and (3) converting rho into a probability vector u through a softmax function, namely deducing the probability u (i) of the ith syndrome in the knowledge graph W by the model D as follows:
Figure BDA0002742733870000231
and taking the cross entropy of u and the 0-1 probability distribution q of the real syndrome result as a loss function of the model D.
The structure of the model D is shown in fig. 6.
(iii) Training to obtain parameters of model D
(iii) training the information vector for each patient p in the training set with the minimization of loss function as the target through the whole process in step (ii). (iii) training all information vectors of each patient p in the training set once according to the process of step (ii), which is called training once on the training set. The whole training process is as follows:
i. the obtained model is stored once per 100 times of training on the training set;
ii. After the primary model is saved, the model is used for primary verification on the verification set to obtain and save the sum of loss functions of all patients in the verification set;
iii, after 100 times of verification, selecting a parameter corresponding to the model with the minimum verification loss function sum as a final parameter of the model D;
iv, performing one test on the test set: and (3) comparing the syndrome with the highest possibility inferred in the step (ii) with the real syndrome for all patients in the test set to obtain the inference accuracy rate of all patients in the test set, and taking the inference accuracy rate as the inference accuracy rate of the model D.
Step (203): and a use method of the model D.
And (3) inputting the information vector of the inpatient p into the model D, finding m elements with the largest numerical values from the probability vector u obtained in the step (202) and (ii), finding m syndromes corresponding to the m elements in the knowledge graph W, outputting the m syndromes, and sequencing according to the probability corresponding to the probability vector u, wherein the larger the numerical value is, the higher the ranking is.
The specific implementation steps of the step (3) are as follows:
next, model G is explained using the first description in the disease description set (denoted as Q), Q _1 as an example.
Step (301): and (3) regarding the ith syndrome H _ i in the information base H in the step (1), taking the symptoms of k patients of the syndrome stored in the H into consideration, calculating the number of people with the symptom Q _1 in the k patients, and recording the number of people as k _ i.
A step (302): and (3) obtaining the number of people k _1, k _2and k _3corresponding to all syndromes through the step (301), sequencing according to the size of k _ i, outputting m syndromes with the largest k _ i, sequencing the m syndromes according to the size of the corresponding k _ i, and ranking the higher k _ i, the higher k _ i is, the higher the ranking is.
Example 2:
an embodiment 2 of the present disclosure provides a medium on which a program is stored, the program implementing the following steps when executed by a processor:
the symptom analysis module analyzes symptom information input by a patient according to a preset knowledge graph based on the syndrome of the symptom;
the inquiry judging module judges whether inquiry is needed or not based on the result of symptom analysis;
if the judgment result is that the inquiry is needed, the inquiry module utilizes the syndrome and the deep learning network model to carry out traditional Chinese medicine inquiry, and obtains the syndrome and/or the syndrome of the patient according to the answer of the patient;
the tongue diagnosis judging module judges whether the tongue diagnosis is combined, if so, the result of the inquiry module enters the combined tongue diagnosis module to obtain the final syndrome and/or syndrome element; otherwise, the result obtained by the inquiry module is used as the final syndrome and/or syndrome element;
when the inquiry judging module judges that the inquiry is not needed or the tongue diagnosis judging module judges that the tongue diagnosis needs to be combined, the tongue diagnosis module is combined to synthesize the symptoms and tongue picture information of the patient based on the syndrome and the deep learning network model to obtain the final syndrome and/or the syndrome.
The detailed steps are the same as those provided in example 1 and are not described again here.
Example 3:
an embodiment 3 of the present disclosure provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, where the processor executes the program to implement the following steps:
the symptom analysis module analyzes symptom information input by a patient according to a preset knowledge graph based on the syndrome of the symptom;
the inquiry judging module judges whether inquiry is needed or not based on the result of symptom analysis;
if the judgment result is that the inquiry is needed, the inquiry module utilizes the syndrome and the deep learning network model to carry out traditional Chinese medicine inquiry, and obtains the syndrome and/or the syndrome of the patient according to the answer of the patient;
the tongue diagnosis judging module judges whether the tongue diagnosis is combined, if so, the result of the inquiry module enters the combined tongue diagnosis module to obtain the final syndrome and/or syndrome element; otherwise, the result obtained by the inquiry module is used as the final syndrome and/or syndrome element;
when the inquiry judging module judges that the inquiry is not needed or the tongue diagnosis judging module judges that the tongue diagnosis needs to be combined, the tongue diagnosis module is combined to synthesize the symptoms and tongue picture information of the patient based on the syndrome and the deep learning network model to obtain the final syndrome and/or the syndrome.
The detailed steps are the same as those provided in example 1 and are not described again here.
As will be appreciated by one of skill in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by a computer program, which may be stored in a computer readable storage medium and executed by a computer to implement the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.

Claims (4)

1. A traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome differentiation and deep learning is characterized by comprising:
a symptom analysis module configured to: analyzing symptom information input by a patient according to a preset knowledge graph based on the syndrome of the symptom;
an interrogation determination module configured to: determining whether an inquiry needs to be made based on the result of the symptom analysis;
an interrogation module configured to: if the judgment result is that the inquiry is needed, the traditional Chinese medicine inquiry is carried out by utilizing the syndrome and the deep learning network model, and the syndrome and/or the syndrome of the patient are/is obtained according to the answer of the patient;
a tongue diagnosis determination module configured to: judging whether tongue diagnosis is combined, if so, entering the result of the inquiry module into a combined tongue diagnosis module to obtain the final syndrome and/or syndrome element; otherwise, the result obtained by the inquiry module is used as the final syndrome and/or syndrome element;
a binding tongue diagnosis module configured to: when the inquiry judging module judges that inquiry is not needed or the tongue diagnosis judging module judges that tongue diagnosis needs to be combined, the final syndrome and/or the syndrome are obtained by integrating the symptoms and tongue picture information of the patient based on the syndrome and the deep learning network model;
the knowledge-graph comprises:
a first knowledge-graph configured to: syndrome corresponding to each symptom;
a second knowledge-graph configured to: categories corresponding to each symptom;
a third knowledge-graph configured to: aiming at typical symptoms corresponding to a certain syndrome in each symptom category;
a fourth knowledge-graph configured to: the syndrome corresponding to each syndrome;
a fifth knowledge-graph configured to: including composite syndrome and ranking;
a sixth knowledge-graph configured to: the corresponding syndrome of each tongue picture;
the symptom analysis module comprises:
transforming symptom descriptions input by the patient into various symptoms stored in a first knowledge graph to obtain a first syndrome set;
aiming at each symptom in the first syndrome set, finding out the disease syndrome and the morbid syndrome corresponding to each symptom in the first knowledge graph, and sequencing all the appeared disease syndromes and morbid syndromes according to the frequency and the sequence of appearance in the symptom;
selecting a plurality of syndrome elements which are ranked most front from the disease syndrome elements and the disease location syndrome elements respectively to form a second syndrome element set;
if all disease symptoms and disease location symptoms corresponding to all symptoms in the first symptom set do not appear in the second symptom set, the symptoms are combined into a third symptom set, and the third symptom set is sorted according to the frequency and the sequence of the appearance in the symptoms;
in the inquiry judging module, whether the evidences in the second evidences set can explain the chief complaints or not is judged, and the evidences contained in the second evidences set can explain a certain symptom, specifically:
for the symptom, if the intersection of the disease syndrome set and the disease location syndrome set corresponding to the symptom in the first knowledge graph and the intersection of the disease syndrome set and the second syndrome set are not empty, the syndrome contained in the second syndrome set is called as that the symptom can be explained, and at this moment, the inquiry is not carried out, and the tongue picture judgment module is entered; otherwise, entering an inquiry module for inquiry;
the inquiry module comprises:
starting with the first disease viseme in the third viseme set;
for any disease symptom element in the third syndrome element set, finding a symptom corresponding to the disease symptom element in the first syndrome element set, finding a category corresponding to the symptom element in the second knowledge graph, finding a typical syndrome element set of the syndrome element in the category in the third knowledge graph, and inquiring whether a certain symptom in the typical syndrome element set appears;
if the answer of the user is yes, selecting the disease nature syndrome, bringing the disease nature syndrome into a second syndrome set, stopping inquiring if the syndrome contained in the new second syndrome set can explain the chief complaint, outputting the second syndrome set, and entering a tongue picture judgment module;
if the evidence element contained in the new second evidence element set can not explain the chief complaint or the answer of the user is no, the next evidence element in the third evidence element set is executed again in the inquiry module;
if the steps in the inquiry module are carried out aiming at all the syndrome elements in the third syndrome element set, the syndrome elements contained in the second syndrome element set can not explain the chief complaints, and the inquiry is carried out by combining the experience of famous old Chinese medicine;
the combined tongue diagnosis module comprises:
converting the tongue picture description of the patient into various tongue pictures stored in a sixth knowledge graph of the traditional Chinese medicine to obtain a tongue picture set;
the second syndrome set obtained by the inquiry module comprises a first syndrome set and a first disease syndrome set;
aiming at each tongue picture in the tongue picture set, finding out the disease property syndrome and disease location syndrome corresponding to each tongue picture in the sixth knowledge graph to obtain a tongue picture syndrome set and obtain a second disease property syndrome set and a second disease location syndrome set;
the intersection of the second syndrome set and the tongue syndrome set obtained by the inquiry module is a first syndrome intersection, the intersection of the first disease syndrome set and the second disease syndrome set is a second syndrome intersection, and the intersection of the first disease syndrome set and the second disease syndrome set is a third syndrome intersection;
if the second syndrome intersection is empty, including:
obtaining a plurality of candidate syndromes through a deep learning network model to obtain a candidate syndrome set, and starting from the first candidate syndrome, entering the next step for any candidate syndrome;
finding out a disease character set corresponding to the candidate syndrome in the fourth knowledge graph, and entering the next step;
if the intersection of the disease nature syndrome set corresponding to the current candidate syndrome and the first disease nature syndrome set is not an empty set, outputting the current candidate syndrome; otherwise, the operation of the previous step is carried out aiming at the next candidate syndrome;
if the disease character set corresponding to all candidate syndromes does not intersect with the first disease character set, outputting a second syndrome set by taking the syndrome deduced from the symptoms as the main;
or,
if the second syndrome intersection contains only one disease syndrome, it includes:
deducing a plurality of candidate syndromes through a deep learning network model to obtain a candidate syndrome set, defining a null set K from a first candidate syndrome, and entering the next step for any candidate syndrome;
if the number of syndromes in K is less than the preset number, finding out a disease character set corresponding to the current candidate syndromes in the fourth knowledge graph, and entering the next step;
if the disease syndrome element set and the second syndrome element intersection have intersection, adding the current candidate syndrome into the set K, and then performing the operation in the previous step aiming at the next candidate syndrome; if the disease syndrome element set and the second syndrome element intersection do not have intersection, the operation in the previous step is carried out aiming at the next candidate syndrome; if the current candidate syndrome is the last syndrome in the candidate syndrome set, entering the next step;
if the number of the syndromes in K reaches a preset value and K is an empty set, outputting a first syndrome intersection, if K is not an empty set, finding a disease property syndrome set corresponding to the current syndrome in K in a fourth knowledge graph, and starting from the first syndrome in K, and entering the next step for each syndrome;
if the intersection of the disease viseme set corresponding to the current candidate viseme in the K is equal to the second viseme, outputting the current candidate viseme in the K, and otherwise, entering the next step;
for each disease viscus which is contained in the disease viscus set corresponding to the current candidate viscus in the K and is not in the second viscus set, finding typical symptoms corresponding to the viscus in a third knowledge graph, inquiring whether a user has one of the typical symptoms, and if the typical symptoms of each disease viscus are answered by the user, outputting the current viscus; otherwise, if the syndrome is not the last syndrome in K, performing the operation in the previous step for the next syndrome in K, and if the syndrome is the last syndrome in K, outputting a first syndrome intersection;
or,
if the second syndrome intersection contains at least two disease syndromes:
according to the ranking of the composite syndromes in the fifth knowledge graph, starting from the first composite syndrome, obtaining a disease property syndrome set corresponding to the current composite syndrome for any one composite syndrome, and entering the next step;
if the second syndrome intersection is equal to the set of disease-related syndromes corresponding to the current compound syndrome, outputting the current compound syndrome; if the second syndrome intersection is not equal to the disease property syndrome set corresponding to the composite syndrome, and the second syndrome intersection is a subset of the disease property syndrome set corresponding to the current composite syndrome, performing the next step;
for each disease viscidity element contained in the disease viscidity element set corresponding to the compound viscidity but not in the second viscidity element set, finding typical symptoms corresponding to the viscidity elements in a third knowledge map, inquiring whether a user has one of the typical symptoms or not, and entering the next step;
if the user answers yes to the typical symptoms of each disease viseme, outputting the current compound syndrome; otherwise, the operation in the previous step is carried out according to the next composite syndrome;
if the intersection of the second syndrome elements is not the subset of the disease syndrome element set corresponding to the composite syndrome, aiming at the next composite syndrome, the operation in the second step is carried out;
and if the composite syndrome cannot be output, outputting the intersection of the first syndrome elements.
2. The system of claim 1, wherein the interrogation module combines the interrogation of the famous old traditional Chinese medicine experience, comprising:
inputting the first evidence set into a deep learning network model, recommending a plurality of candidate syndromes through the deep learning network model, and carrying out the following operations on any candidate syndrome from the first candidate syndrome:
finding out a disease condition element set corresponding to the current syndrome in a fourth knowledge graph, stopping inquiry if the disease condition elements in the disease condition element set are all in the second syndrome set, outputting the current syndrome, and entering a tongue picture judgment module; otherwise, composing the disease visemes in the disease viseme set but not in the second viseme set into a fourth viseme set;
starting from the first disease property element in the fourth element set, inquiring each disease property element whether a user has a certain symptom in the typical element set, if so, outputting the current syndrome and entering a tongue picture judgment module; if the user answers no to a certain symptom element, the first two steps of operation are carried out for the next syndrome;
if any one of the candidate syndromes cannot be output by the inquiry module in combination with the inquiry operation of the famous traditional Chinese medicine experience, a second syndrome set is output and the tongue picture judgment module is entered.
3. The intelligent traditional Chinese medicine interrogation tongue diagnosis comprehensive system based on syndrome and deep learning of claim 2, wherein if the results of the interrogation judgment module and the interrogation module are certain syndromes, the obtained results are directly used as final syndromes; and if not, substituting the second evidential collection output by the inquiry judging module and the inquiry module into the combined tongue diagnosis module.
4. An electronic device comprising a memory, a processor and a program stored on the memory and running on the processor, wherein the processor implements the steps in the system according to any one of claims 1-3 when executing the program.
CN202011155816.3A 2020-10-26 2020-10-26 Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning Active CN112216383B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011155816.3A CN112216383B (en) 2020-10-26 2020-10-26 Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011155816.3A CN112216383B (en) 2020-10-26 2020-10-26 Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning

Publications (2)

Publication Number Publication Date
CN112216383A CN112216383A (en) 2021-01-12
CN112216383B true CN112216383B (en) 2023-02-21

Family

ID=74056479

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011155816.3A Active CN112216383B (en) 2020-10-26 2020-10-26 Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning

Country Status (1)

Country Link
CN (1) CN112216383B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114496231B (en) * 2022-02-16 2024-03-26 平安科技(深圳)有限公司 Knowledge graph-based constitution identification method, device, equipment and storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108182262A (en) * 2018-01-04 2018-06-19 华侨大学 Intelligent Answer System construction method and system based on deep learning and knowledge mapping
CN108877928A (en) * 2018-05-31 2018-11-23 平安医疗科技有限公司 Patient information acquisition method, device, computer equipment and storage medium
CN110459321A (en) * 2019-08-20 2019-11-15 山东众阳健康科技集团有限公司 A kind of aided diagnosis of traditional Chinese medicine system based on card element
CN110838368A (en) * 2019-11-19 2020-02-25 广州西思数字科技有限公司 Robot active inquiry method based on traditional Chinese medicine clinical knowledge graph

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102222153A (en) * 2010-01-27 2011-10-19 洪文学 Quantitative dialectical diagnostic method for Chinese medicine machine interrogation
JP2019536137A (en) * 2016-10-25 2019-12-12 コーニンクレッカ フィリップス エヌ ヴェKoninklijke Philips N.V. Knowledge diagnosis based clinical diagnosis support
CN108986912A (en) * 2018-07-12 2018-12-11 北京三医智慧科技有限公司 Chinese medicine stomach trouble tongue based on deep learning is as information intelligent processing method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108182262A (en) * 2018-01-04 2018-06-19 华侨大学 Intelligent Answer System construction method and system based on deep learning and knowledge mapping
CN108877928A (en) * 2018-05-31 2018-11-23 平安医疗科技有限公司 Patient information acquisition method, device, computer equipment and storage medium
CN110459321A (en) * 2019-08-20 2019-11-15 山东众阳健康科技集团有限公司 A kind of aided diagnosis of traditional Chinese medicine system based on card element
CN110838368A (en) * 2019-11-19 2020-02-25 广州西思数字科技有限公司 Robot active inquiry method based on traditional Chinese medicine clinical knowledge graph

Also Published As

Publication number Publication date
CN112216383A (en) 2021-01-12

Similar Documents

Publication Publication Date Title
US10854335B2 (en) Computer aided medical method and medical system for medical prediction
CN110598786B (en) Neural network training method, semantic classification method and semantic classification device
CN113688255A (en) Knowledge graph construction method based on Chinese electronic medical record
CN113707299A (en) Auxiliary diagnosis method and device based on inquiry session and computer equipment
CN115050442B (en) Disease category data reporting method and device based on mining clustering algorithm and storage medium
CN117747087A (en) Training method of large inquiry model, inquiry method and device based on large inquiry model
Mondal et al. Wme: Sense, polarity and affinity based concept resource for medical events
CN114343577A (en) Cognitive function evaluation method, terminal device, and computer-readable storage medium
CN112216383B (en) Traditional Chinese medicine intelligent inquiry tongue diagnosis comprehensive system based on syndrome element and deep learning
CN113871020A (en) Health management method and system after critical illness diagnosis based on AI machine learning
CN112071431B (en) Clinical path automatic generation method and system based on deep learning and knowledge graph
CN114300127A (en) Method, device, equipment and storage medium for inquiry processing
CN117252265A (en) Human-computer interaction method, system, medium and equipment based on interpretable machine learning
CN112259232A (en) VTE risk automatic evaluation system based on deep learning
CN110047569B (en) Method, device and medium for generating question-answer data set based on chest radiography report
CN112052327A (en) Method of knowledge point mastering condition analysis system
CN116629385A (en) GPT model optimization method and device
CN109859813A (en) A kind of entity modification word recognition method and device
CN112598202B (en) Test question difficulty evaluation method and device, storage medium and computing equipment
Mastnak Coherence size and confidence range: two new parameters in psycho-cardiology
TW202331737A (en) Electronic medical record data analysis system and electronic medical record data analysis method
CN114520053A (en) Medical information processing method, system, terminal and storage medium
Aguirre-Celis et al. From Words to Sentences & Back: Characterizing Context-dependent Meaning Representations in the Brain
Romero-Gómez et al. Natural Language Processing Approach for Learning Process Analysis in a Bioinformatics Course
TWI852643B (en) Artificial intelligence facial emotion recognition system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CP01 Change in the name or title of a patent holder
CP01 Change in the name or title of a patent holder

Address after: 12 / F, building 1, Aosheng building, 1166 Xinluo street, hi tech Zone, Jinan City, Shandong Province

Patentee after: Zhongyang Health Technology Group Co.,Ltd.

Address before: 12 / F, building 1, Aosheng building, 1166 Xinluo street, hi tech Zone, Jinan City, Shandong Province

Patentee before: SHANDONG MSUNHEALTH TECHNOLOGY GROUP Co.,Ltd.