CN116386805A - Intelligent guided diagnosis report generation method - Google Patents
Intelligent guided diagnosis report generation method Download PDFInfo
- Publication number
- CN116386805A CN116386805A CN202310393556.0A CN202310393556A CN116386805A CN 116386805 A CN116386805 A CN 116386805A CN 202310393556 A CN202310393556 A CN 202310393556A CN 116386805 A CN116386805 A CN 116386805A
- Authority
- CN
- China
- Prior art keywords
- patient
- report
- degree
- information
- patients
- 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.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 176
- 238000000034 method Methods 0.000 title claims abstract description 56
- 230000003993 interaction Effects 0.000 claims abstract description 114
- 208000024891 symptom Diseases 0.000 claims abstract description 66
- 230000002452 interceptive effect Effects 0.000 claims abstract description 39
- 230000008569 process Effects 0.000 claims abstract description 22
- 239000013589 supplement Substances 0.000 claims abstract description 21
- 238000001514 detection method Methods 0.000 claims description 38
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 18
- 239000013598 vector Substances 0.000 claims description 15
- 201000010099 disease Diseases 0.000 claims description 14
- 238000011084 recovery Methods 0.000 claims description 13
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000010521 absorption reaction Methods 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 8
- 238000011156 evaluation Methods 0.000 claims description 8
- 230000007246 mechanism Effects 0.000 claims description 7
- 238000013210 evaluation model Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 6
- 230000000007 visual effect Effects 0.000 claims description 5
- 230000035790 physiological processes and functions Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000008901 benefit Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000009469 supplementation Effects 0.000 claims 1
- 238000001356 surgical procedure Methods 0.000 claims 1
- 208000002193 Pain Diseases 0.000 description 4
- 208000035475 disorder Diseases 0.000 description 4
- 230000003872 anastomosis Effects 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 230000008520 organization Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 208000019695 Migraine disease Diseases 0.000 description 1
- 210000001015 abdomen Anatomy 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 206010001902 amaurosis Diseases 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000000481 breast Anatomy 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 206010027599 migraine Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002741 palatine tonsil Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 230000008961 swelling Effects 0.000 description 1
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
- G16H15/00—ICT specially adapted for medical reports, e.g. generation or transmission thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/242—Dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- Y—GENERAL 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Public Health (AREA)
- Animal Behavior & Ethology (AREA)
- Primary Health Care (AREA)
- Medical Informatics (AREA)
- Epidemiology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The application provides an intelligent guided diagnosis report generation method, which comprises the following steps: acquiring user data to establish a patient personal information system, and comparing the interactive data with a corpus to judge the type of the patient; the system sends related inquiry sentences to carry out interactive guidance supplement on the patient interactive data; counting user interaction process data, and analyzing the acceptance degree of a patient for a guided diagnosis report; judging the learning level of the patient and the understanding degree of medical knowledge according to the context information of the user interaction data; judging complaint symptoms according to the description information after interaction guidance supplement and by combining a patient personal information system; matching a target diagnosis guiding department according to the complaint symptoms; matching the hospital according to the acquired patient information; generating a personalized structured lead report according to the refined patient type; and optimizing the output content and typesetting of the guided report.
Description
Technical Field
The invention relates to the technical field of information, in particular to an intelligent guided diagnosis report generation method.
Background
With the development and progress of society, the medical technology field related to life health of people has been rapidly developed, and a general medical procedure is to first register in a hospital and then diagnose a specific disease. When the patient does not know the illness state of the patient, the patient does not know what department should be hung, and the patient can go to the hospital to make a diagnosis relatively blindly, and a diagnosis guiding report mode is derived to guide the patient to make a diagnosis for the patient to make an autonomous diagnosis; the prior guided diagnosis report has high specialization degree and more medical terms, is difficult to understand for the patient with low cultural degree, and leads the patient to have insufficient knowledge of own illness state, so that the patient neglects the illness state and selects the wrong department of medical treatment. Or misunderstand the diagnosis guiding result and can not be well matched with the result after diagnosis guiding to seek medical attention. Under serious conditions, the opportunity of visit is delayed, the treatment cost is increased, even the life and the health of the patient are endangered, and the irrecoverable loss is caused for the patient; therefore, how to make the user know the own illness state deeply and quickly and accurately see the diagnosis guiding report, so that the user can see the diagnosis according to the diagnosis guiding scheme, and the user can grasp the diagnosis opportunity, thereby having important practical significance and theoretical research value.
Disclosure of Invention
The invention provides a method for generating an intelligent guided diagnosis report, which mainly comprises the following steps:
acquiring user data to establish a patient personal information system, and comparing the interactive data with a corpus to judge the type of the patient; the system sends related inquiry sentences to carry out interactive guidance supplement on the patient interactive data; counting user interaction process data, and analyzing the acceptance degree of a patient for a guided diagnosis report; judging the learning level of the patient and the understanding degree of the medical knowledge according to the context information of the user interaction data, wherein the judging the learning level of the patient and the understanding degree of the medical knowledge according to the context information of the user interaction data specifically comprises the following steps: based on personal information of the patient and a patient knowledge questionnaire, respectively analyzing the academic levels of the two types of patients, and based on the understanding degree of the patient to the medical terms by the user interaction data, performing medical knowledge graph quality detection on the patient interaction data information; judging complaint symptoms according to the description information after interaction guidance supplement and by combining a patient personal information system; matching a target diagnosis guiding department according to the complaint symptoms; matching the hospital according to the acquired patient information; generating a personalized structured lead report according to the refined patient type, wherein the generating the personalized structured lead report according to the refined patient type specifically comprises the following steps: generating a civilian report for patients with low academic levels, and generating a specialized guided diagnosis report for patients with high academic levels; and optimizing the output content and typesetting of the guided report.
Further optionally, the acquiring the user data establishes a personal information system of the patient, and the discriminating the type of the patient according to the comparison between the interactive data and the corpus includes:
interactive record data acquired by a patient himself on-line diagnosis guiding service platform or a hospital off-line diagnosis guiding service machine; wherein the interaction record data comprises interaction data, interaction process data and patient personal information; the patient personal information includes name, date of birth, residence information, academy and medical history; the interactive record data is subjected to data cleaning by utilizing an ETL mode, and missing data, repeated data and error data are deleted; constructing the patient personal information system by using the cleaned data, and checking the authenticity of the patient personal information through a hospital system authentication interface under the condition that the patient authorizes and agrees; the patient types comprise two types of high cultural level and low cultural level; the dialogue scene of the corpus specifically refers to a doctor-patient inquiry dialogue scene; firstly, counting according to acquired user interaction data; constructing a patient type list, wherein the patient type list is established according to the occurrence frequency of the medical knowledge entities or a scoring mechanism of the relation among the medical knowledge entities in the user interaction data; the relation among the medical knowledge entities is expressed by entity sequence connection entities, if the entity has no association relation with other entities, the relation is recorded as 1, if the entity has association relation with other entities, the score is recorded as 2, and the like; the scoring mechanism is s=fi×ci, fi represents occurrence frequency of the ith entity, and Ci represents association relation score of the ith entity; comprehensively analyzing the score values of different patients based on the statistics of the medical knowledge entities and the relation between the medical knowledge entities; and inputting the score value of the user dialogue scene and the acquired user click, browsing speed and using habit data into a deep neural network, and carrying out two classification on the type of the patient according to the model, wherein the two classification is high in cultural degree/low in cultural degree.
Further optionally, the interactive guidance supplementing of the patient interaction data by the system sending related query sentences includes:
the input of the patient interaction data is divided into a plurality of sections of text input and a plurality of sections of voice input; the voice text conversion module is used for converting the multi-section voice input into complete multi-section text section by section; integrating text information through text input or voice conversion; extracting disorder keywords from the patient interaction data information, wherein the number of the keywords can be one or more, and the keywords exist or are similar in a disorder database; extracting keywords in the obtained text information, identifying the intention of the patient in the input sentence, and obtaining an actual intention target in the patient interaction through model analysis and processing; the actual intended goal of the patient is the patient's complaint symptoms; by utilizing the fusion medical knowledge term and the context information, meaning expressed in the patient interaction message is analyzed, and further, intention information of a patient is obtained, wherein the intention information of the patient comprises two states of 'unclear' and 'symptom description'; the method comprises the steps of performing interactive guidance supplement on missing information of interactive expression of a patient by analyzing the vocabulary number, grammar structure and grammar morphology of each text information of the patient; for patients with unclear intention information expression, interactive guidance supplement is focused on the expression information of the patients, and the patients of the type are guided to describe more comprehensive disease description information.
Further optionally, the statistical user interaction process data, analyzing the patient's acceptance of the lead report includes:
the user interaction process data comprise user click, picture viewing times and browsing speed; the user click is used for analyzing the user interaction capability, the click counter is inserted into the guide operation page to record the user click times, the user interaction capability is taken as the percentage of the user browsing times to the average browsing times of the whole user, and the report form refers to animation demonstration for explaining the professional pictures; the number of times of checking the picture is used for analyzing the attention degree of the image content, the number of times of checking the picture by the user is recorded by inserting a checking picture counter in a diagnosis guiding operation page, and the attention degree of the user is taken as the percentage of the number of times of checking the picture by the user to the average number of times of checking the whole user; the browsing speed is used for analyzing the text content absorption degree, the browsing speeds of two types of users are measured and calculated through the automatic page browsing duration acquisition function, and the browsing speed is taken as the text content absorption degree of the users according to the percentage of the browsing speed of the users to the average speed of the whole users; establishing an evaluation model of the acceptance degree of the guided diagnosis report form, namely
E=w1×j+w2×p+w3×t, where E represents specialized and diversified triage report acceptance degree triage report acceptance degrees, W1, W2, W3 represent different weights, and w1+w2+w3=1, J represents the interactive ability of the user, P represents the attention degree of the user, and T represents the text content absorption degree of the user; the weight of W1, W2 and W3 is used as the distribution probability of the contents of pictures, characters and interaction forms in the guided diagnosis report, and the contents of the whole guided diagnosis report are planned according to an evaluation model of the acceptance degree of the guided diagnosis report form, wherein the contents of the guided diagnosis report comprise civiliation and specialization; calculating the acceptance degree of all patients of different types on specialized and diversified diagnosis guiding reports, and simultaneously calculating the average acceptance degree of the whole user; if the acceptance of the guided diagnosis report of a patient is higher than the average acceptance of the user, the patient is more capable of accepting specialized and diversified guided diagnosis reports than other patients.
Further optionally, the determining the learning level and the understanding degree of the medical knowledge of the patient according to the context information of the user interaction data includes:
the learning level of the patient is judged by the context information of the user interaction data; the patients with low academic grade refer to users with high, medium and lower academic grade, and the rest users are classified as patients with high academic grade; patients with high academic levels have high desire to learn medical knowledge; for patients of the high-grade type of academic, the patient does not necessarily know about the relevant medical knowledge, although having a relatively high triage report acceptance capability; aiming at the patients with high academic grade and low understanding degree of the medical terms, the noun explanation of the specialized medical terms is added into the guided diagnosis report, so that the patients can know own illness state and specialized information deeply; comprising the following steps: based on personal information of the patient and a patient knowledge questionnaire, respectively analyzing the academic grades of the two types of patients; analyzing the degree of understanding of the medical term by the patient based on the user interaction data; performing medical knowledge graph quality detection on the patient interaction data information;
The learning grade of two types of patients is respectively analyzed based on personal information of the patients and combined with a patient knowledge questionnaire, and the learning grade comprises the following specific steps:
firstly, based on the learning of personal information of a patient, combining with the questionnaire data of the knowledge of the patient, and further judging the learning level of the patient; the text information obtained by the knowledge questionnaire reflects the accumulated times of wrong questions and knowledge seeking scores; respectively carrying out frequency statistics on the text information, and combining the academy of the personal information of the patient as the academy grade evaluation index data, namely index= { academy, error question accumulation times and knowledge finding desire score }, wherein index represents the academy grade evaluation index; determining a training set and a testing set according to the text information, taking the training set of the text information as the characteristic input, training an ANN neural network learning model, and obtaining a discrimination report of the user learning level according to the final output result of the neural network learning model.
The analyzing the understanding degree of the patient to the medical term based on the user interaction data specifically comprises the following steps:
the understanding degree of the medical term is obtained through quality detection of a medical knowledge graph; the method comprises the steps of carrying out refinement classification on patients with high academic grade, and dividing the patients into two categories of high understanding degree of medical terms and low understanding degree of the medical terms; in addition, for patients with high academic levels and low understanding of medical terms, it is considered to add noun interpretation of specialized medical terms to their lead report.
The medical knowledge graph quality detection of the patient interaction data information specifically comprises the following steps:
firstly, presetting a medical knowledge graph; the relationship between entities is often stored in the form of triples in the medical knowledge graph, and the specific structure is as follows: < entity-relationship-entity >, < entity-attribute value >, the medical knowledge graph is established by the logical relationship connection between the entity and other entities; the quality detection of the medical knowledge graph is considered by three aspects of triad structure combination specification detection, triad standardization detection and knowledge consistency detection, wherein the quality detection of the medical knowledge graph refers to the steps of extracting entity of patient interaction data information and judging whether a disease described by a patient meets medical theoretical knowledge or not; the detection of the triad structure combination specification refers to calculating the degree of each entity, wherein the degree of the entity refers to the input degree and the output degree, whether structural errors exist or not is judged according to the degree of the entity, the structural errors comprise isolated nodes and isolated nodes of the entity, and if the entity is the isolated node, the entity is deleted; the isolated node refers to a node with the entity degree of 0, the isolated node refers to a node with the entity degree of 1, and the entity degree of the associated node of the isolated node is 1; the standardized detection of the triplets refers to combining symptoms by adopting a method of entity alignment; the entity alignment is to calculate an entity similarity score by adopting the Lychnical ratio, the same character ratio and the semantic vector distance, wherein the entity similarity score is greater than 0.8, and the entity similarity is judged; the knowledge consistency detection is carried out by combining rules, wherein the rules comprise physiological state association rules and symptom association rules.
Further optionally, the determining the complaint symptom according to the description information after the interaction guidance and the personal information system of the patient includes:
according to the description information after interaction guidance supplement as input, acquiring the existing medical history data and the past medical history data by combining a patient personal information system on the basis of the interaction guidance supplement, preprocessing the data to obtain a training data set, taking symptoms as labels, and using the training set for a BERT model; obtaining the current intention category of the user according to the BERT classification model; when the current intention category of the user is symptom description, using a symptom dictionary as a custom dictionary of a word segmentation tool, and carrying out word segmentation and part-of-speech recognition according to the symptom dictionary and current input information; extracting keywords in the current input information according to parts of speech, matching the keywords with symptom terms, if the matching is successful, finding out complaint symptoms of a patient, otherwise, training a BERT model by adopting an unsupervised SimCSE method to obtain a sentence vectorization representation model, and respectively inputting the symptom terms and the current input information into the sentence vectorization representation model to obtain respective corresponding semantic vectors; and calculating the similarity between the semantic vector of the symptom term and the semantic vector of the current input information, and determining the complaint symptom of the user according to the calculation result.
Further optionally, the matching the target diagnosis guiding department according to the complaint symptoms includes:
determining a first candidate department set based on the main complaint symptoms, and constructing a weighted directed graph taking the main complaint symptoms and each first candidate department as nodes; setting personalized weights of each node, wherein the nodes represent departments, the personalized weights are determined by a hierarchical analysis method, and iterative computation is carried out on the personalized weights by adopting an edge weight and PageRank algorithm to obtain the score of each node; normalizing each score to obtain a normalized score, and calculating a first base index of department score distribution in a first waiting department set according to the normalized score; if the first base index is smaller than the second base index, recommending the first candidate diagnosis room as a diagnosis guiding target department.
Further optionally, the matching the hospital according to the acquired patient information includes:
acquiring residence information contained in user basic data of the interactive record data; acquiring residence information and local map information according to a patient personal information system, and calculating a navigation route from the residence of the patient to each hospital; acquiring visit notes and predicted navigation route time of each hospital, comparing route spending reaching each hospital, and screening hospitals meeting a route spending threshold; on the basis of considering route expense, comprehensive recovery rate, sub-average diagnosis cost and patient satisfaction information of each hospital are obtained; obtaining patient medical record information from a patient personal information system, and analyzing and knowing the average demand of a patient on budget and service quality; acquiring recovery rate, sub-average diagnosis cost, patient satisfaction information and visit notes of each hospital within a set route overhead threshold; comprehensively considering the patient demands, and recommending the hospitals which are most in line with the actual demands to the user by combining the information of the recovery rate, the sub-average diagnosis cost and the patient satisfaction of all hospitals within the set route spending threshold; and adding the navigation route, the recovery rate, the charge level, the patient satisfaction degree and the visit notes to a diagnosis guiding chart, and returning the diagnosis guiding chart to the client.
Further optionally, the generating a personalized structured lead report according to the refined patient type includes:
the refined patient type refers to that the patients with high academic grade are subjected to two categories of high understanding degree of the medical terms and low understanding degree of the medical terms, the patients with high academic grade are not required to know the desire to know the medical terms, and noun interpretation of the technical terms is added in the guided report of the patients with the type; according to the statistical user interaction process data, the acceptance degree of two types of users for the guided diagnosis report is known; the structured guided diagnosis report is an electronic report with a certain template, and the content and the form of the report comprise picture content, text content and an animation demonstration part; comprising the following steps: generating a civilian report for a patient of a low-grade type of the academic; generating a specialized diagnosis guiding report aiming at a patient with a high academic grade;
the generation of the civilian report for the patient with the low academic grade comprises the following steps:
users with low academic levels, including both patients with low and high medical term understanding; a civilian guided diagnosis report is adopted for patients with low academic grade; aiming at the patients with low academic grade, judging that the patients cannot understand the medical term in the interaction process, and adopting a mode of generating a civilian guided diagnosis report; the civilian diagnosis guiding report is adopted, so that a patient with low school grade can quickly obtain popular and easily understood information, and the user is persuaded to make a diagnosis in time, and the template of the diagnosis guiding report is adjusted according to different interaction capacities.
The generation of the specialized diagnosis guiding report for the patient with high academic grade specifically comprises the following steps:
users with high academic levels include patients with low understanding of medical terms and patients with high understanding of medical terms; a specialized diagnosis guiding report is adopted for patients with high academic grade; aiming at the patients with high academic grade and low understanding degree of medical terms, a specialized diagnosis guiding report capable of guiding rapid and accurate diagnosis is adopted; aiming at a patient with high academic grade and low understanding degree of medical terms, in the interaction process, judging that the understanding degree of the patient to the medical terms is not high, so that the patient cannot know the own illness state deeply and neglects the illness state, and generating a specialized diagnosis guiding report containing medical term explanation; considering that the patient with high academic grade has high acceptance degree of the report, the advantage of combining the electronic report in the specialized guided diagnosis report is added with reasonable picture content and animation demonstration on the basis of text content.
Further optionally, the optimizing the output content and typesetting of the guided report includes:
the optimization direction of the intelligent guided diagnosis report comprises the paragraph number, the professional term interpretation paragraph, the term quantity, the picture specialty and typesetting; according to the acceptance degree of the guided diagnosis report by the user, the academic level of the user and the understanding degree of the medical term by the user, two structural guided diagnosis report templates of civilian and specialized are adjusted; aiming at the patients with low academic grade, adopting civilian guided diagnosis report, optimizing the paragraph number and term quantity of text content according to the acceptance degree of the guided diagnosis report, and performing simple and straight-white animation demonstration on specialized pictures; aiming at the patients with high academic grade, a specialized diagnosis guiding report is adopted, and according to the understanding capability of the patients to the medical terms, the explanation paragraphs of the specialized terms are optimized mainly for the patients with low understanding degree of the medical terms, and the optimization direction is concise and visual for the patients with high understanding degree of the medical terms.
The technical scheme provided by the embodiment of the invention can have the following beneficial effects:
according to the invention, the cultural degree of the user and the understanding of the medical terms can be judged according to the interaction of the user, the accurate diagnosis guiding can be carried out according to the information of the user, and the targeted structured diagnosis guiding report is generated according to the difference of the interaction expression modes, so that patients with different cultural degrees can know the own illness state, people with high cultural degrees can know more specialized information, people with low cultural degrees can also quickly obtain popular and easily understood information, and the user is convinced to carry out timely diagnosis.
Drawings
FIG. 1 is a flow chart of a method for generating an intelligent guided diagnosis report according to the present invention.
FIG. 2 is a flow chart of a method of analyzing patient interaction data according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and specifically described below with reference to the drawings in the embodiments of the present invention. The described embodiments are only a few embodiments of the present invention.
The method for generating the intelligent guided diagnosis report in the embodiment specifically comprises the following steps:
step 101, acquiring user data, establishing a patient personal information system, and comparing and judging the type of the patient according to the interactive data and the corpus.
Interactive record data acquired by a patient himself on-line diagnosis guiding service platform or a hospital off-line diagnosis guiding service machine; wherein the interaction record data comprises interaction data, interaction process data and patient personal information; the patient personal information includes name, date of birth, residence information, academy and medical history; the interactive record data is subjected to data cleaning by utilizing an ETL mode, and missing data, repeated data and error data are deleted; and constructing the patient personal information system by using the cleaned data, and checking the authenticity of the patient personal information through a hospital system authentication interface under the condition that the patient agrees with authorization. The patient types comprise two types of high cultural level and low cultural level; the dialogue scene of the corpus specifically refers to a doctor-patient inquiry dialogue scene; firstly, counting according to acquired user interaction data; constructing a patient type list, wherein the patient type list is established according to the occurrence frequency of the medical knowledge entities or a scoring mechanism of the relation among the medical knowledge entities in the user interaction data; the relation among the medical knowledge entities is expressed by entity sequence connection entities, if the entity has no association relation with other entities, the relation is recorded as 1, if the entity has association relation with other entities, the score is recorded as 2, and the like; the scoring mechanism is s=fi×ci, fi represents occurrence frequency of the ith entity, and Ci represents association relation score of the ith entity; comprehensively analyzing the score values of different patients based on the statistics of the medical knowledge entities and the relation between the medical knowledge entities; inputting the score value of the user dialogue scene and the acquired user click, browsing speed and using habit data into a deep neural network, and carrying out two classifications on the type of the patient according to the model, wherein the two classifications are high in cultural degree/low in cultural degree; for example, patient interaction record data are respectively obtained through a hospital offline diagnosis guiding service machine and a hospital online diagnosis guiding service platform, if a plurality of pieces of obtained user basic data belong to the same patient, only one piece of relevant data is reserved by deleting repeated name data and date of birth data in an ETL (Extract-Transform-Load) mode, and other data information is summarized. For example, after data processing, interactive data in patient interaction record data of a patient A, a patient B, a patient C and a patient D are obtained, and scoring values of different patients are calculated according to a scoring mechanism; assuming that the patient A does not use medical knowledge terms in the interactive expression through analysis, the patient B uses the medical knowledge terms and has a comprehensive score of 85 points, the patient C uses the medical knowledge terms and has a comprehensive score of 45 points, and the patient D uses the medical knowledge terms and has a comprehensive score of 65 points; the patient score value is input into a CNN network model together with the acquired user click, browsing speed and using habit data, and the model classification obtains a conclusion: patient A and patient C are marked as low in cultural degree, patient B and patient D are marked as high in cultural degree, and finally the marked states of all users are recorded in a patient type list, namely { patient name, patient cultural degree }.
And 102, performing interactive guidance supplement on the patient interaction data by sending related query sentences through the system.
The input of the patient interaction data is divided into a plurality of sections of text input and a plurality of sections of voice input; the voice text conversion module is used for converting the multi-section voice input into complete multi-section text section by section; integrating text information through text input or voice conversion; extracting disorder keywords from the patient interaction data information, wherein the number of the keywords can be one or more, and the keywords exist or are similar in a disorder database; extracting keywords in the obtained text information, identifying the intention of the patient in the input sentence, and obtaining an actual intention target in the patient interaction through model analysis and processing; the actual intended goal of the patient is the patient's complaint symptoms; by utilizing the fusion medical knowledge term and the context information, meaning expressed in the patient interaction message is analyzed, and further, intention information of a patient is obtained, wherein the intention information of the patient comprises two states of 'unclear' and 'symptom description'; the method comprises the steps of performing interactive guidance supplement on missing information of interactive expression of a patient by analyzing the vocabulary number, grammar structure and grammar morphology of each text information of the patient; aiming at the patient with unclear intention information expression, the expression information is subjected to interactive guidance supplement to guide the patient to describe more comprehensive disease description information; for example, predicting vocabulary by the BERT model depending on the context information, and accurately describing semantic information of sentences or even sections of chapter level; the last step of output of the system and text input of text or voice conversion of a user are used as the input of a BERT model, in an initial state, no last step of system output can be replaced by blank characters, and the current user input is text input information of text input or voice conversion in a doctor-patient interaction system, which generally refers to a consultation state of the user and takes the actual intention of the current user as a label; if the last step of input of the system is ' tonsil red swelling ' and the user input information is ' unclear ', the input of the model is ' CLS ' and the right upper abdomen distending pain is ' SEP ' is unclear ', the intention classification label is ' information ambiguous ', and when the conditions occur, interactive guidance is carried out to supplement more complete intention information for the user; if the previous step of input of the system is "right epigastric distending pain" and the user input information is "yes", then input of the model is "[ CLS ] right epigastric distending pain is" [ SEP ], the intention classification label is "symptom description", and the output vector of [ CLS ] is generally selected as input of the full connection layer, and prediction probabilities of various categories are obtained through MLP.
And 103, counting user interaction process data, and analyzing the acceptance degree of the patient for the guided diagnosis report.
The user interaction process data comprise user click, picture viewing times and browsing speed; the user click is used for analyzing the user interaction capability, the click counter is inserted into the guide operation page to record the user click times, the user interaction capability is taken as the percentage of the user browsing times to the average browsing times of the whole user, and the report form refers to animation demonstration for explaining the professional pictures; the number of times of checking the picture is used for analyzing the attention degree of the image content, the number of times of checking the picture by the user is recorded by inserting a checking picture counter in a diagnosis guiding operation page, and the attention degree of the user is taken as the percentage of the number of times of checking the picture by the user to the average number of times of checking the whole user; the browsing speed is used for analyzing the text content absorption degree, the browsing speeds of two types of users are measured and calculated through the automatic page browsing duration acquisition function, and the browsing speed is taken as the text content absorption degree of the users according to the percentage of the browsing speed of the users to the average speed of the whole users; establishing an evaluation model of the acceptance degree of the guided diagnosis report form, namely
E=w1×j+w2×p+w3×t, where E represents specialized and diversified triage report acceptance degree triage report acceptance degrees, W1, W2, W3 represent different weights, and w1+w2+w3=1, J represents the interactive ability of the user, P represents the attention degree of the user, and T represents the text content absorption degree of the user; the weight of W1, W2 and W3 is used as the distribution probability of the contents of pictures, characters and interaction forms in the guided diagnosis report, and the contents of the whole guided diagnosis report are planned according to an evaluation model of the acceptance degree of the guided diagnosis report form, wherein the contents of the guided diagnosis report comprise civiliation and specialization; calculating the acceptance degree of all patients of different types on specialized and diversified diagnosis guiding reports, and simultaneously calculating the average acceptance degree of the whole user; if the acceptance degree of the guided diagnosis report of a certain patient is higher than the average acceptance degree of the user, the patient is more capable of accepting specialized and diversified guided diagnosis reports than other patients; for example: setting acceptance weights W1, W2 and W3 of the guided diagnosis report form of the patient A as 0.4,0.1,0.5, wherein the probability of enriching the report form is 0.4, the probability of inserting the picture content is 0.5, the probability of describing the text content is 0.8, and the acceptance of the guided diagnosis report form of the patient A is 0.6; traversing all users and calculating that the average acceptance of all user guided report forms is 0.5, the patient A is more capable of accepting specialized and diversified guided reports.
And 104, judging the learning level of the patient and the understanding degree of medical knowledge according to the context information of the user interaction data.
The learning level of the patient is judged by the context information of the user interaction data; the patients with low academic grade refer to users with high, medium and lower academic grade, and the rest users are classified as patients with high academic grade; patients with high academic levels have high desire to learn medical knowledge; for patients of the high-grade type of academic, the patient does not necessarily know about the relevant medical knowledge, although having a relatively high triage report acceptance capability; aiming at the patients with high academic grade and low understanding degree of the medical terms, the noun explanation of the specialized medical terms is added into the guided diagnosis report, so that the patients can know own illness state and specialized information deeply; for example: judging the learning level of the patient and the familiarity degree of the patient with medical knowledge according to the context information of the user interaction data, providing a civilian diagnosis guiding report for the patient A if the cultural degree of the patient A is junior middle school, enabling the content to be popular and easy to understand, aiming at enabling the patient A to be able to see the diagnosis in time, analyzing according to the interaction information data if the cultural degree of the patient B is the family, knowing that the familiarity degree of the patient B with medical terms is not high, providing a specialized diagnosis guiding report for the patient B in order to enable the patient B to know the state of illness deeply, and adding medical term explanation into the report to help the patient to be better familiar with medical knowledge.
The academic levels of the two types of patients are analyzed separately based on the patient personal information in combination with the patient knowledge questionnaire.
Firstly, based on the learning of personal information of a patient, combining with the questionnaire data of the knowledge of the patient, and further judging the learning level of the patient; the text information obtained by the knowledge questionnaire reflects the accumulated times of wrong questions and knowledge seeking scores; respectively carrying out frequency statistics on the text information, and combining the academy of the personal information of the patient as the academy grade evaluation index data, namely index= { academy, error question accumulation times and knowledge finding desire score }, wherein index represents the academy grade evaluation index; determining a training set and a testing set according to the text information, taking the training set of the text information as the characteristic input, training an ANN neural network learning model, and obtaining a discrimination report of the user learning level according to the final output result of the neural network learning model; for example: the evaluation system of the patient's academic level is the master and above (0.9-1), the family (0.7-0.89), the senior citizen (0.4-0.69), the junior middle school and below (0-0.39), and the input layer of the ANN neural network comprises 3 indexes: the learning, the number of accumulated wrong questions and knowledge seeking, namely n=3, the output layer is the final learning level evaluation result, namely m=1, and the hidden layer is 3 layers; after the interaction data is trained by the ANN neural network, the learning grade values of the patients A, B, C and D are 0.85,0.65,0.92,0.33 respectively, the learning grade of the patient A is the family, the learning grade of the patient B is the middle school, the learning grade of the patient C is the master and above, and the learning grade of the patient D is the junior middle school and below.
The patient's degree of understanding of the medical terms is analyzed based on the user interaction data.
The understanding degree of the medical term is obtained through quality detection of a medical knowledge graph; the method comprises the steps of carrying out refinement classification on patients with high academic grade, and dividing the patients into two categories of high understanding degree of medical terms and low understanding degree of the medical terms; in addition, for patients with high academic grade and low understanding degree of medical terms, noun explanation of specialized medical terms is considered to be added into the guided report; for example: and detecting the similarity of the medical knowledge graph according to the user interaction data, and calculating that the similarity between the user interaction data information of the patient A and the medical knowledge graph is greater than a preset threshold value of 0.8, wherein the patient A is considered to have high understanding degree on the medical terms, and if the similarity is lower than the preset threshold value, the patient A is considered to have low understanding degree on the medical terms.
And detecting the quality of the medical knowledge graph of the patient interaction data information.
Firstly, presetting a medical knowledge graph; the relationship between entities is often stored in the form of triples in the medical knowledge graph, and the specific structure is as follows: < entity-relationship-entity >, < entity-attribute value >, the medical knowledge graph is established by the logical relationship connection between the entity and other entities; the quality detection of the medical knowledge graph is considered by three aspects of triad structure combination specification detection, triad standardization detection and knowledge consistency detection, wherein the quality detection of the medical knowledge graph refers to the steps of extracting entity of patient interaction data information and judging whether a disease described by a patient meets medical theoretical knowledge or not; the detection of the triad structure combination specification refers to calculating the degree of each entity, wherein the degree of the entity refers to the input degree and the output degree, whether structural errors exist or not is judged according to the degree of the entity, the structural errors comprise isolated nodes and isolated nodes of the entity, and if the entity is the isolated node, the entity is deleted; the isolated node refers to a node with the entity degree of 0, the isolated node refers to a node with the entity degree of 1, and the entity degree of the associated node of the isolated node is 1; the standardized detection of the triplets refers to combining symptoms by adopting a method of entity alignment; the entity alignment is to calculate an entity similarity score by adopting the Lychnical ratio, the same character ratio and the semantic vector distance, wherein the entity similarity score is greater than 0.8, and the entity similarity is judged; the knowledge consistency detection is carried out by combining rules, wherein the rules comprise physiological state association rules and symptom association rules; for example: according to the user interaction data information, extracting the relation between the entities to construct triples, sequentially carrying out triple structure combination specification detection, triple standardization detection and knowledge consistency detection on the triples, and carrying out identification feedback operation on the detected abnormal entities; the entity name corresponding to the entity category of 'symptom' is 'eye blacking' degree of 1, is an unassociated entity, is an isolated node and needs to be deleted; disease-symptoms, symptoms of the disease "migraine" are "blacking before eye" and are "recurrent", but there is no "recurrent" in the character vocabulary in the blacking before eye of the symptoms, which is a structural error, which requires to complement the character in the physical character vocabulary; the symptoms of "amaurosis" and "blackness before eyes" are the same in meaning, and only one standardized entity expression exists in the same kind of entity, so that the two symptoms are combined into one; symptoms of "male breast distending pain", incorrect association of physiological states, need careful feedback detection; according to the existing medical knowledge graph, measuring the characteristics of the entity, such as semantic vectors, relation vectors and the like, and calculating the similarity score of the entity, namely the quality detection of the medical knowledge graph, wherein the higher the score is, the higher the quality of the medical knowledge graph is.
And 105, judging the complaint symptoms by combining the personal information system of the patient according to the description information after the interaction guidance supplement.
According to the description information after interaction guidance supplement as input, acquiring the existing medical history data and the past medical history data by combining a patient personal information system on the basis of the interaction guidance supplement, preprocessing the data to obtain a training data set, taking symptoms as labels, and using the training set for a BERT model; obtaining the current intention category of the user according to the BERT classification model; when the current intention category of the user is symptom description, using a symptom dictionary as a custom dictionary of a word segmentation tool, and carrying out word segmentation and part-of-speech recognition according to the symptom dictionary and current input information; extracting keywords in the current input information according to parts of speech, matching the keywords with symptom terms, if the matching is successful, finding out complaint symptoms of a patient, otherwise, training a BERT model by adopting an unsupervised SimCSE method to obtain a sentence vectorization representation model, and respectively inputting the symptom terms and the current input information into the sentence vectorization representation model to obtain respective corresponding semantic vectors; and calculating the similarity between the semantic vector of the symptom term and the semantic vector of the current input information, and determining the complaint symptom of the user according to the calculation result.
And 106, matching the target diagnosis guiding department according to the complaint symptoms.
Determining a first candidate department set based on the main complaint symptoms, and constructing a weighted directed graph taking the main complaint symptoms and each first candidate department as nodes; setting personalized weights of each node, wherein the nodes represent departments, the personalized weights are determined by a hierarchical analysis method, and iterative computation is carried out on the personalized weights by adopting an edge weight and PageRank algorithm to obtain the score of each node; normalizing each score to obtain a normalized score, and calculating a first base index of department score distribution in a first waiting department set according to the normalized score; if the first base index is smaller than the second base index, recommending the first candidate diagnosis room as a diagnosis guiding target department; for example: the method comprises the steps that 10 departments are selected in a certain hospital, the characteristics of the 'department anastomosis disease' are divided into two groups according to a threshold value of 0.2, 6 departments are matched, doctors in 5 departments are idle, doctors in 1 department are busy, a data set is divided into two parts of D1 and D2, and the characteristics of the 'department anastomosis disease' are marked as A; in D1, 6 departments have a threshold value greater than 0.2, wherein 5 doctors in the departments are idle, and 1 doctor in the department is busy; the threshold value of 4 departments in the D2 is less than 0.2, wherein doctors in 4 departments are busy; the coefficient of the data set D1, G (D1) =2× (5/6) × (1/6) =10/36, and the duty cycle of D1 is 6/10; the coefficient of the data set D2, G (D2) =2× (0/4) × (4/4) =0, and the duty cycle weight of D1 is 4/10; according to the formula: g (D, A) = (6/10) = (10/36) + (4/10) ×0, calculating to obtain a base coefficient G (D, A) of 0.17 under the characteristic of 'department anastomosis disease', assuming that the second base index is 0.48, comparing to obtain that the first base index is smaller than the second base index, and selecting the first candidate diagnosis room as a target diagnosis room for diagnosis.
Step 107, matching the hospital according to the acquired patient information.
Acquiring residence information contained in user basic data of the interactive record data; acquiring residence information and local map information according to a patient personal information system, and calculating a navigation route from the residence of the patient to each hospital; acquiring visit notes and predicted navigation route time of each hospital, comparing route spending reaching each hospital, and screening hospitals meeting a route spending threshold; on the basis of considering route expense, comprehensive recovery rate, sub-average diagnosis cost and patient satisfaction information of each hospital are obtained; obtaining patient medical record information from a patient personal information system, and analyzing and knowing the average demand of a patient on budget and service quality; acquiring recovery rate, sub-average diagnosis cost, patient satisfaction information and visit notes of each hospital within a set route overhead threshold; comprehensively considering the patient demands, and recommending the hospitals which are most in line with the actual demands to the user by combining the information of the recovery rate, the sub-average diagnosis cost and the patient satisfaction of all hospitals within the set route spending threshold; adding the navigation route, the recovery rate, the charge level, the patient satisfaction degree and the visit notes into a diagnosis guiding chart, and returning the diagnosis guiding chart to a client; for example: the department corresponding to the complaint symptoms of the patient A is arranged in the hospital A, the hospital B and the hospital C, residence information of personal information of the patient A in the system is fetched, a route overhead threshold value is set to be 10 minutes, routes from residence of the patient A to three hospitals are respectively calculated, the routes are sequentially 10 minutes, 5 minutes and 20 minutes, and the hospital A and the hospital B are compared and known to meet the requirements; the medical record information of the patient A in the system is fetched, the demand of the patient A is analyzed and known to be recovered as early as possible, the budget is within 300 yuan, and the required service quality is high; the recovery rate of corresponding departments of three hospitals is 32%,25% and 28% in sequence, the charge of the three hospitals is 359 yuan, 224 yuan and 291 yuan in sequence, and the service satisfaction degree of the three hospitals is satisfied in sequence, and generally, the comprehensive comparison recommends a hospital B to a patient A.
Step 108, generating a personalized structured lead report according to the refined patient type.
The refined patient type refers to that the patients with high academic grade are subjected to two categories of high understanding degree of the medical terms and low understanding degree of the medical terms, the patients with high academic grade are not required to know the desire to know the medical terms, and noun interpretation of the technical terms is added in the guided report of the patients with the type; according to the statistical user interaction process data, the acceptance degree of two types of users for the guided diagnosis report is known; the structured guided diagnosis report is an electronic report with a certain template, and the content and the form of the report comprise picture content, text content and an animation demonstration part; for example: patient A is farmer, patient B is the medical student of family, patient C is the non-medical student of the master, when the guided report is the indiscriminate structured guided report, if patient A receives the guided report with high degree of specialization and more medical terms, because it is difficult to understand the report, neglect the illness state or choose the wrong consulting room, in serious cases, delay the opportunity to visit, increase the treatment cost, even endanger the life and health of patient, cause irrecoverable loss to patient; the patient C is not aware of the medical terms, so that the patient is not aware of the disease condition of the patient deeply, or misunderstands the diagnosis guiding result, and the patient C cannot well cooperate with the diagnosis guiding result to carry out medical treatment, and an explanation paragraph of the medical terms is necessary to be added in the diagnosis guiding report; in order to make patient A better understand his own illness state and make a diagnosis in time, a popular and easily understood diagnosis guiding report is generated for him.
A civilian report is generated for a patient of a low-grade type of academic.
Users with low academic levels, including both patients with low and high medical term understanding; a civilian guided diagnosis report is adopted for patients with low academic grade; aiming at the patients with low academic grade, judging that the patients cannot understand the medical term in the interaction process, and adopting a mode of generating a civilian guided diagnosis report; the method comprises the steps that a civilian diagnosis guiding report is adopted, so that a patient with low academic grade can quickly obtain popular and easily understood information, a user is persuaded to make a diagnosis in time, and a template of the diagnosis guiding report is adjusted according to different interaction capacities; for example: aiming at the patients with low academic level, the patients with high acceptance degree of pictures and animation demonstration and low acceptance degree of a large number of characters are obtained according to statistics of the user interaction process data; patients with low academic grades have differences in the acceptance degree of the guided report, and for patients with strong interaction ability in the class, the form of rich reports is considered; the academic grade of the patient A is judged to be lower, but the patient A is high in acceptance of the guided diagnosis report through analysis, and the guided diagnosis report with rich styles is high in acceptance, so that the patient A is considered to be added with rich animation demonstration, the space of pictures and characters in the report is increased, and the purpose of enabling the patient A to obtain clear illness state information through the guided diagnosis report is achieved, and the purpose of convincing the patient A to make a timely visit is achieved.
Generating a specialized diagnosis guiding report aiming at a patient with a high academic grade.
Users with high academic levels include patients with low understanding of medical terms and patients with high understanding of medical terms; a specialized diagnosis guiding report is adopted for patients with high academic grade; aiming at the patients with high academic grade and low understanding degree of medical terms, a specialized diagnosis guiding report capable of guiding rapid and accurate diagnosis is adopted; aiming at a patient with high academic grade and low understanding degree of medical terms, in the interaction process, judging that the understanding degree of the patient to the medical terms is not high, so that the patient cannot know the own illness state deeply and neglects the illness state, and generating a specialized diagnosis guiding report containing medical term explanation; considering that the patient with high academic grade has high acceptance degree of the report, combining the advantage of the electronic report in the specialized guided diagnosis report, adding reasonable picture content and animation demonstration on the basis of text content; for example: the patient A is low in understanding degree of medical terms through analysis, so that the patient is not deep enough in understanding the disease condition of the patient, or misunderstanding the diagnosis guiding result, and the patient can not well cooperate with the diagnosis guiding result to see the doctor, so that the purposes of accurately guiding the doctor and not delaying the diagnosis time are achieved, and the interpretation section of the professional medical terms is considered to be added aiming at the specialized diagnosis guiding report of the patient A.
And step 109, optimizing the output content and typesetting of the guided report.
The optimization direction of the intelligent guided diagnosis report comprises the paragraph number, the professional term interpretation paragraph, the term quantity, the picture specialty and typesetting; according to the acceptance degree of the guided diagnosis report by the user, the academic level of the user and the understanding degree of the medical term by the user, two structural guided diagnosis report templates of civilian and specialized are adjusted; aiming at the patients with low academic grade, adopting civilian guided diagnosis report, optimizing the paragraph number and term quantity of text content according to the acceptance degree of the guided diagnosis report, and performing simple and straight-white animation demonstration on specialized pictures; aiming at the patients with high academic grade, a specialized diagnosis guiding report is adopted, and according to the understanding capability of the patients to the medical terms, the explanation paragraphs of the specialized terms are optimized mainly for the patients with low understanding degree of the medical terms, and the optimization direction is simple and visual for the patients with high understanding degree of the medical terms; for example, patient A has a low grade of academic and a high degree of acceptance of the triage report, patient B has a high grade of academic and a low degree of acceptance of the triage report, patient C has a high grade of academic and a high degree of understanding of medical terms, and patient D has a high degree of cultural and a low degree of understanding of medical terms; reducing the number of paragraphs and the term amount of patient A and patient B; aiming at a patient A, a civilian diagnosis guiding report is adopted, and a simple and visual typesetting is adopted, so that in view of high acceptance of the diagnosis guiding report, specialized pictures are added into the report and are matched with simple and straight-white animation demonstration, so that the patient A is helped to know own illness state; aiming at a patient B, a civilian diagnosis guiding report is adopted, and a simple and visual typesetting is adopted, so that the patient B is persuaded to visit in time by using popular and easily understood language in view of low acceptance of the diagnosis guiding report without adding specialized pictures; aiming at a patient C, a specialized diagnosis guiding report is adopted, the picture expertise is improved, and the typesetting style of a composition organization is adopted; the specialized guided diagnosis report is adopted for the patient D, the typesetting style of the organization is adopted, the number of paragraphs and the term quantity of the report are properly increased, and term interpretation paragraphs are additionally added.
The foregoing is merely illustrative of the structures of this invention and various modifications, additions and substitutions for those skilled in the art can be made to the described embodiments without departing from the scope of the invention or from the scope of the invention as defined in the accompanying claims.
Claims (10)
1. An intelligent guided diagnosis report generating method, which is characterized by comprising the following steps:
acquiring user data to establish a patient personal information system, and comparing the interactive data with a corpus to judge the type of the patient; the system sends related inquiry sentences to carry out interactive guidance supplement on the patient interactive data; counting user interaction process data, and analyzing the acceptance degree of a patient for a guided diagnosis report; judging the learning level of the patient and the understanding degree of the medical knowledge according to the context information of the user interaction data, wherein the judging the learning level of the patient and the understanding degree of the medical knowledge according to the context information of the user interaction data specifically comprises the following steps: based on personal information of the patient and a patient knowledge questionnaire, respectively analyzing the academic levels of the two types of patients, and based on the understanding degree of the patient to the medical terms by the user interaction data, performing medical knowledge graph quality detection on the patient interaction data information; judging complaint symptoms according to the description information after interaction guidance supplement and by combining a patient personal information system; matching a target diagnosis guiding department according to the complaint symptoms; matching the hospital according to the acquired patient information; generating a personalized structured lead report according to the refined patient type, wherein the generating the personalized structured lead report according to the refined patient type specifically comprises the following steps: generating a civilian report for patients with low academic levels, and generating a specialized guided diagnosis report for patients with high academic levels; and optimizing the output content and typesetting of the guided report.
2. The method of claim 1, wherein the acquiring user data establishes a patient personal information system, and discriminating the type of the patient based on the interactive data compared with the corpus comprises:
interactive record data acquired by a patient himself on-line diagnosis guiding service platform or a hospital off-line diagnosis guiding service machine; wherein the interaction record data comprises interaction data, interaction process data and patient personal information; the patient personal information includes name, date of birth, residence information, academy and medical history; the interactive record data is subjected to data cleaning by utilizing an ETL mode, and missing data, repeated data and error data are deleted; constructing the patient personal information system by using the cleaned data, and checking the authenticity of the patient personal information through a hospital system authentication interface under the condition that the patient authorizes and agrees; the patient types comprise two types of high cultural level and low cultural level; the dialogue scene of the corpus specifically refers to a doctor-patient inquiry dialogue scene; firstly, counting according to acquired user interaction data; constructing a patient type list, wherein the patient type list is established according to the occurrence frequency of the medical knowledge entities or a scoring mechanism of the relation among the medical knowledge entities in the user interaction data; the relation among the medical knowledge entities is expressed by entity sequence connection entities, if the entity has no association relation with other entities, the relation is recorded as 1, if the entity has association relation with other entities, the score is recorded as 2, and the like; the scoring mechanism is s=fi×ci, fi represents occurrence frequency of the ith entity, and Ci represents association relation score of the ith entity; comprehensively analyzing the score values of different patients based on the statistics of the medical knowledge entities and the relation between the medical knowledge entities; and inputting the score value of the user dialogue scene and the acquired user click, browsing speed and using habit data into a deep neural network, and carrying out two classification on the type of the patient according to the model, wherein the two classification is high in cultural degree/low in cultural degree.
3. The method of claim 1, wherein the interactive guided supplementation of patient interaction data by the system sending relevant query sentences comprises:
the input of the patient interaction data is divided into a plurality of sections of text input and a plurality of sections of voice input; the voice text conversion module is used for converting the multi-section voice input into complete multi-section text section by section; integrating text information through text input or voice conversion; extracting disorder keywords from the patient interaction data information, wherein the number of the keywords can be one or more, and the keywords exist or are similar in a disorder database; extracting keywords in the obtained text information, identifying the intention of the patient in the input sentence, and obtaining an actual intention target in the patient interaction through model analysis and processing; the actual intended goal of the patient is the patient's complaint symptoms; by utilizing the fusion medical knowledge term and the context information, meaning expressed in the patient interaction message is analyzed, and further, intention information of a patient is obtained, wherein the intention information of the patient comprises two states of 'unclear' and 'symptom description'; the method comprises the steps of performing interactive guidance supplement on missing information of interactive expression of a patient by analyzing the vocabulary number, grammar structure and grammar morphology of each text information of the patient; for patients with unclear intention information expression, interactive guidance supplement is focused on the expression information of the patients, and the patients of the type are guided to describe more comprehensive disease description information.
4. The method of claim 1, wherein the statistical user interaction process data analyzing patient acceptance of the lead report comprises:
the user interaction process data comprise user click, picture viewing times and browsing speed; the user click is used for analyzing the user interaction capability, the click counter is inserted into the guide operation page to record the user click times, the user interaction capability is taken as the percentage of the user browsing times to the average browsing times of the whole user, and the report form refers to animation demonstration for explaining the professional pictures; the number of times of checking the picture is used for analyzing the attention degree of the image content, the number of times of checking the picture by the user is recorded by inserting a checking picture counter in a diagnosis guiding operation page, and the attention degree of the user is taken as the percentage of the number of times of checking the picture by the user to the average number of times of checking the whole user; the browsing speed is used for analyzing the text content absorption degree, the browsing speeds of two types of users are measured and calculated through the automatic page browsing duration acquisition function, and the browsing speed is taken as the text content absorption degree of the users according to the percentage of the browsing speed of the users to the average speed of the whole users; establishing an evaluation model of the acceptance degree of the guided report form, namely E=W1×J+W2×P+W3×T, wherein E represents specialization and diversification of the guided report acceptance degree, W1, W2 and W3 represent different weights, W1+W2+W3=1, J represents the interaction capability of a user, P represents the attention degree of the user, and T represents the text content absorption degree of the user; the weight of W1, W2 and W3 is used as the distribution probability of the contents of pictures, characters and interaction forms in the guided diagnosis report, and the contents of the whole guided diagnosis report are planned according to an evaluation model of the acceptance degree of the guided diagnosis report form, wherein the contents of the guided diagnosis report comprise civiliation and specialization; calculating the acceptance degree of all patients of different types on specialized and diversified diagnosis guiding reports, and simultaneously calculating the average acceptance degree of the whole user; if the acceptance of the guided diagnosis report of a patient is higher than the average acceptance of the user, the patient is more capable of accepting specialized and diversified guided diagnosis reports than other patients.
5. The method of claim 1, wherein the determining the patient's level of learning and the degree of understanding of medical knowledge based on the contextual information of the user interaction data comprises:
the learning level of the patient is judged by the context information of the user interaction data; the patients with low academic grade refer to users with high, medium and lower academic grade, and the rest users are classified as patients with high academic grade; patients with high academic levels have high desire to learn medical knowledge; for patients of the high-grade type of academic, the patient does not necessarily know about the relevant medical knowledge, although having a relatively high triage report acceptance capability; aiming at the patients with high academic grade and low understanding degree of the medical terms, the noun explanation of the specialized medical terms is added into the guided diagnosis report, so that the patients can know own illness state and specialized information deeply; comprising the following steps: based on personal information of the patient and a patient knowledge questionnaire, respectively analyzing the academic grades of the two types of patients; analyzing the degree of understanding of the medical term by the patient based on the user interaction data; performing medical knowledge graph quality detection on the patient interaction data information;
the learning grade of two types of patients is respectively analyzed based on personal information of the patients and combined with a patient knowledge questionnaire, and the learning grade comprises the following specific steps:
Firstly, based on the learning of personal information of a patient, combining with the questionnaire data of the knowledge of the patient, and further judging the learning level of the patient; the text information obtained by the knowledge questionnaire reflects the accumulated times of wrong questions and knowledge seeking scores; respectively carrying out frequency statistics on the text information, and combining the academy of the personal information of the patient as the academy grade evaluation index data, namely index= { academy, error question accumulation times and knowledge finding desire score }, wherein index represents the academy grade evaluation index; determining a training set and a testing set according to the text information, taking the training set of the text information as the characteristic input, training an ANN neural network learning model, and obtaining a discrimination report of the user learning level according to the final output result of the neural network learning model;
the analyzing the understanding degree of the patient to the medical term based on the user interaction data specifically comprises the following steps:
the understanding degree of the medical term is obtained through quality detection of a medical knowledge graph; the method comprises the steps of carrying out refinement classification on patients with high academic grade, and dividing the patients into two categories of high understanding degree of medical terms and low understanding degree of the medical terms; in addition, for patients with high academic grade and low understanding degree of medical terms, noun explanation of specialized medical terms is considered to be added into the guided report;
The medical knowledge graph quality detection of the patient interaction data information specifically comprises the following steps:
firstly, presetting a medical knowledge graph; the relationship between entities is often stored in the form of triples in the medical knowledge graph, and the specific structure is as follows: < entity-relationship-entity >, < entity-attribute value >, the medical knowledge graph is established by the logical relationship connection between the entity and other entities; the quality detection of the medical knowledge graph is considered by three aspects of triad structure combination specification detection, triad standardization detection and knowledge consistency detection, wherein the quality detection of the medical knowledge graph refers to the steps of extracting entity of patient interaction data information and judging whether a disease described by a patient meets medical theoretical knowledge or not; the detection of the triad structure combination specification refers to calculating the degree of each entity, wherein the degree of the entity refers to the input degree and the output degree, whether structural errors exist or not is judged according to the degree of the entity, the structural errors comprise isolated nodes and isolated nodes of the entity, and if the entity is the isolated node, the entity is deleted; the isolated node refers to a node with the entity degree of 0, the isolated node refers to a node with the entity degree of 1, and the entity degree of the associated node of the isolated node is 1; the standardized detection of the triplets refers to combining symptoms by adopting a method of entity alignment; the entity alignment is to calculate an entity similarity score by adopting the Lychnical ratio, the same character ratio and the semantic vector distance, wherein the entity similarity score is greater than 0.8, and the entity similarity is judged; the knowledge consistency detection is carried out by combining rules, wherein the rules comprise physiological state association rules and symptom association rules.
6. The method of claim 1, wherein the determining complaint symptoms in conjunction with the patient's personal information system based on the interactive guidance post-replenishment descriptive information comprises:
according to the description information after interaction guidance supplement as input, acquiring the existing medical history data and the past medical history data by combining a patient personal information system on the basis of the interaction guidance supplement, preprocessing the data to obtain a training data set, taking symptoms as labels, and using the training set for a BERT model; obtaining the current intention category of the user according to the BERT classification model; when the current intention category of the user is symptom description, using a symptom dictionary as a custom dictionary of a word segmentation tool, and carrying out word segmentation and part-of-speech recognition according to the symptom dictionary and current input information; extracting keywords in the current input information according to parts of speech, matching the keywords with symptom terms, if the matching is successful, finding out complaint symptoms of a patient, otherwise, training a BERT model by adopting an unsupervised SimCSE method to obtain a sentence vectorization representation model, and respectively inputting the symptom terms and the current input information into the sentence vectorization representation model to obtain respective corresponding semantic vectors; and calculating the similarity between the semantic vector of the symptom term and the semantic vector of the current input information, and determining the complaint symptom of the user according to the calculation result.
7. The method of claim 1, wherein the matching target guided surgery based on complaint symptoms comprises:
determining a first candidate department set based on the main complaint symptoms, and constructing a weighted directed graph taking the main complaint symptoms and each first candidate department as nodes; setting personalized weights of each node, wherein the nodes represent departments, the personalized weights are determined by a hierarchical analysis method, and iterative computation is carried out on the personalized weights by adopting an edge weight and PageRank algorithm to obtain the score of each node; normalizing each score to obtain a normalized score, and calculating a first base index of department score distribution in a first waiting department set according to the normalized score; if the first base index is smaller than the second base index, recommending the first candidate diagnosis room as a diagnosis guiding target department.
8. The method of claim 1, wherein the matching the hospital according to the acquired patient information comprises:
acquiring residence information contained in user basic data of the interactive record data; acquiring residence information and local map information according to a patient personal information system, and calculating a navigation route from the residence of the patient to each hospital; acquiring visit notes and predicted navigation route time of each hospital, comparing route spending reaching each hospital, and screening hospitals meeting a route spending threshold; on the basis of considering route expense, comprehensive recovery rate, sub-average diagnosis cost and patient satisfaction information of each hospital are obtained; obtaining patient medical record information from a patient personal information system, and analyzing and knowing the average demand of a patient on budget and service quality; acquiring recovery rate, sub-average diagnosis cost, patient satisfaction information and visit notes of each hospital within a set route overhead threshold; comprehensively considering the patient demands, and recommending the hospitals which are most in line with the actual demands to the user by combining the information of the recovery rate, the sub-average diagnosis cost and the patient satisfaction of all hospitals within the set route spending threshold; and adding the navigation route, the recovery rate, the charge level, the patient satisfaction degree and the visit notes to a diagnosis guiding chart, and returning the diagnosis guiding chart to the client.
9. The method of claim 1, wherein the generating a personalized structured lead report from the refined patient type comprises:
the refined patient type refers to that the patients with high academic grade are subjected to two categories of high understanding degree of the medical terms and low understanding degree of the medical terms, the patients with high academic grade are not required to know the desire to know the medical terms, and noun interpretation of the technical terms is added in the guided report of the patients with the type; according to the statistical user interaction process data, the acceptance degree of two types of users for the guided diagnosis report is known; the structured guided diagnosis report is an electronic report with a certain template, and the content and the form of the report comprise picture content, text content and an animation demonstration part; comprising the following steps: generating a civilian report for a patient of a low-grade type of the academic; generating a specialized diagnosis guiding report aiming at a patient with a high academic grade;
the generation of the civilian report for the patient with the low academic grade comprises the following steps:
users with low academic levels, including both patients with low and high medical term understanding; a civilian guided diagnosis report is adopted for patients with low academic grade; aiming at the patients with low academic grade, judging that the patients cannot understand the medical term in the interaction process, and adopting a mode of generating a civilian guided diagnosis report; the method comprises the steps that a civilian diagnosis guiding report is adopted, so that a patient with low academic grade can quickly obtain popular and easily understood information, a user is persuaded to make a diagnosis in time, and a template of the diagnosis guiding report is adjusted according to different interaction capacities;
The generation of the specialized diagnosis guiding report for the patient with high academic grade specifically comprises the following steps:
users with high academic levels include patients with low understanding of medical terms and patients with high understanding of medical terms; a specialized diagnosis guiding report is adopted for patients with high academic grade; aiming at the patients with high academic grade and low understanding degree of medical terms, a specialized diagnosis guiding report capable of guiding rapid and accurate diagnosis is adopted; aiming at a patient with high academic grade and low understanding degree of medical terms, in the interaction process, judging that the understanding degree of the patient to the medical terms is not high, so that the patient cannot know the own illness state deeply and neglects the illness state, and generating a specialized diagnosis guiding report containing medical term explanation; considering that the patient with high academic grade has high acceptance degree of the report, the advantage of combining the electronic report in the specialized guided diagnosis report is added with reasonable picture content and animation demonstration on the basis of text content.
10. The method of claim 1, wherein the optimizing output content and layout of the lead report comprises:
the optimization direction of the intelligent guided diagnosis report comprises the paragraph number, the professional term interpretation paragraph, the term quantity, the picture specialty and typesetting; according to the acceptance degree of the guided diagnosis report by the user, the academic level of the user and the understanding degree of the medical term by the user, two structural guided diagnosis report templates of civilian and specialized are adjusted; aiming at the patients with low academic grade, adopting civilian guided diagnosis report, optimizing the paragraph number and term quantity of text content according to the acceptance degree of the guided diagnosis report, and performing simple and straight-white animation demonstration on specialized pictures; aiming at the patients with high academic grade, a specialized diagnosis guiding report is adopted, and according to the understanding capability of the patients to the medical terms, the explanation paragraphs of the specialized terms are optimized mainly for the patients with low understanding degree of the medical terms, and the optimization direction is concise and visual for the patients with high understanding degree of the medical terms.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310393556.0A CN116386805A (en) | 2023-04-13 | 2023-04-13 | Intelligent guided diagnosis report generation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310393556.0A CN116386805A (en) | 2023-04-13 | 2023-04-13 | Intelligent guided diagnosis report generation method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116386805A true CN116386805A (en) | 2023-07-04 |
Family
ID=86964155
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310393556.0A Pending CN116386805A (en) | 2023-04-13 | 2023-04-13 | Intelligent guided diagnosis report generation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116386805A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116821375A (en) * | 2023-08-29 | 2023-09-29 | 之江实验室 | Cross-institution medical knowledge graph representation learning method and system |
CN117271804A (en) * | 2023-11-21 | 2023-12-22 | 之江实验室 | Method, device, equipment and medium for generating common disease feature knowledge base |
CN117316412A (en) * | 2023-11-09 | 2023-12-29 | 中南大学湘雅医院 | Ophthalmic diagnosis guiding control method and system based on voice prompt |
-
2023
- 2023-04-13 CN CN202310393556.0A patent/CN116386805A/en active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116821375A (en) * | 2023-08-29 | 2023-09-29 | 之江实验室 | Cross-institution medical knowledge graph representation learning method and system |
CN116821375B (en) * | 2023-08-29 | 2023-12-22 | 之江实验室 | Cross-institution medical knowledge graph representation learning method and system |
CN117316412A (en) * | 2023-11-09 | 2023-12-29 | 中南大学湘雅医院 | Ophthalmic diagnosis guiding control method and system based on voice prompt |
CN117271804A (en) * | 2023-11-21 | 2023-12-22 | 之江实验室 | Method, device, equipment and medium for generating common disease feature knowledge base |
CN117271804B (en) * | 2023-11-21 | 2024-03-01 | 之江实验室 | Method, device, equipment and medium for generating common disease feature knowledge base |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20210233658A1 (en) | Identifying Relevant Medical Data for Facilitating Accurate Medical Diagnosis | |
CN107403068B (en) | Merge the intelligence auxiliary way of inquisition and system of clinical thinking | |
US10331659B2 (en) | Automatic detection and cleansing of erroneous concepts in an aggregated knowledge base | |
CN111753098A (en) | Teaching method and system based on cross-media dynamic knowledge graph | |
CN116386805A (en) | Intelligent guided diagnosis report generation method | |
Teng et al. | Automatic medical code assignment via deep learning approach for intelligent healthcare | |
US20200320363A1 (en) | Artificial intelligence advisory systems and methods for vibrant constitutional guidance | |
Li et al. | Reliable medical diagnosis from crowdsourcing: Discover trustworthy answers from non-experts | |
CN110277167A (en) | The Chronic Non-Communicable Diseases Risk Forecast System of knowledge based map | |
WO2023178971A1 (en) | Internet registration method, apparatus and device for seeking medical advice, and storage medium | |
CN113569023A (en) | Chinese medicine question-answering system and method based on knowledge graph | |
CN115293161A (en) | Reasonable medicine taking system and method based on natural language processing and medicine knowledge graph | |
CN116992002A (en) | Intelligent care scheme response method and system | |
Sileo et al. | DiscSense: Automated semantic analysis of discourse markers | |
CN113254609B (en) | Question-answering model integration method based on negative sample diversity | |
Labbé et al. | ChatGPT for phenotypes extraction: one model to rule them all? | |
Li et al. | Approach of intelligence question-answering system based on physical fitness knowledge graph | |
CN111930908A (en) | Answer recognition method and device based on artificial intelligence, medium and electronic equipment | |
CN115312186B (en) | Auxiliary screening system for diabetic retinopathy | |
CN116719840A (en) | Medical information pushing method based on post-medical-record structured processing | |
Li et al. | Finding similar medical questions from question answering websites | |
CN114496231A (en) | Constitution identification method, apparatus, equipment and storage medium based on knowledge graph | |
CN114281959B (en) | Statement processing method, device, equipment, medium and computer program product | |
Feng et al. | Extracting meaningful correlations among heterogeneous datasets for medical question answering with domain knowledge | |
CN117038026A (en) | Recommendation method, electronic equipment and medium for hospital specialists |
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 |