CN112786194A - Medical image diagnosis guide inspection system, method and equipment based on artificial intelligence - Google Patents

Abstract

Translated fromChinese

本发明公开的基于人工智能的医学影像导诊导检系统，包括：数据资源模块用于创建、整合慢病知识图谱、影像图谱、影像专业诊断资料等医学影像数据资源；匹配模块用于构建匹配模型，将病例信息输入匹配模型进行匹配，得到多路融合粗筛召回结果集；导诊模块用于将粗筛召回结果集提取患者业务相关联特征作为数据语料，送入精排模型，进行训练、筛选，得到科普性导诊、知识、风险预测等结果集；导检模块用于将粗筛召回结果集提取医生业务相关联特征作为数据语料，送入精排模型，进行训练、筛选，得到专业性诊疗方案、诊断提醒等导检服务结果集。可以实现针对医学影像类的检查项目、就诊方案与服务等AI辅助推荐与问答咨询，辅助患者自查与医生诊断。

The artificial intelligence-based medical image guidance system for diagnosis and inspection disclosed in the present invention includes: a data resource module for creating and integrating medical image data resources such as chronic disease knowledge maps, image maps, and imaging professional diagnostic data; a matching module for constructing matching Model, input the case information into the matching model for matching, and obtain a multi-channel fusion coarse screening recall result set; the guide module is used to extract the patient's business-related features from the coarse screening recall result set as data corpus, and send it to the fine sorting model for training. , screening, and obtain the result sets of popular science guide, knowledge, risk prediction, etc.; the guide module is used to extract the relevant features of the doctor's business from the coarse screening recall result set as the data corpus, and send it to the fine sorting model for training, screening, and get The result set of guided inspection services such as professional diagnosis and treatment plans, diagnosis reminders, etc. It can realize AI-assisted recommendation and question-and-answer consultation for medical imaging inspection items, medical treatment plans and services, and assist patients in self-examination and doctor diagnosis.

Description

Medical image diagnosis guide inspection system, method and equipment based on artificial intelligence

Technical Field

The invention relates to the technical field of intelligent medical treatment, in particular to a medical image diagnosis guide and inspection system, method, equipment and medium based on artificial intelligence.

Background

During the visit, many patients often experience confusion in choosing a department; medical knowledge is too professional, clinical manifestations of diseases are complex, and a large number of diseases have similar symptoms, so that a patient is confused or even wrong when choosing a department; at present, manual diagnosis guide not only consumes time and labor, but also greatly influences medical service efficiency and patient experience of seeking medical advice. The existing intelligent diagnosis guide, examination guide and case structured category products are relatively comprehensive in technical function points and service application, but are incomplete in the field subdivision aspect of service objects, and image application is lacked.

Disclosure of Invention

Aiming at the defects in the prior art, the embodiment of the invention provides a medical image diagnosis guide inspection system, a method, equipment and a medium based on artificial intelligence, in particular relates to the application of an AI question-answering system, a knowledge map, risk prediction and recommendation system in the technical field of artificial intelligence.

In a first aspect, an embodiment of the present invention provides a medical image guided diagnosis and guidance system based on artificial intelligence, including: an inquiry module, a case generation module, a data resource module, a matching module, a diagnosis guide module and a detection guide module,

the inquiry module is used for collecting basic information, chief complaint information and AI inquiry exchange information of the patient;

the case generation module is used for generating structured cases from the patient information and pushing the cases to the patient end and the doctor end respectively;

the data resource module is used for creating and integrating a chronic disease knowledge map, an image examination map and medical image data resources of image professional diagnosis data;

the matching module is used for constructing a matching model according to an integrated matching algorithm, a matching rule and a knowledge map retrieval rule, extracting features from medical image data resources, manufacturing an image data corpus, training the matching model by adopting a machine learning method, inputting case information into the matching model for matching, and obtaining a multi-path fusion coarse screening recall result set;

the diagnosis guide module is used for extracting characteristics associated with the patient service from the multi-path fusion coarse screening recall result set, sending the characteristics as supplementary linguistic data into the first fine ranking model, training, sequencing and screening to obtain diagnosis guide information, disease knowledge and risk prediction result sets, and sending a recommendation result to a patient end;

and the inspection guide module is used for extracting characteristics associated with doctor service from the multi-path fusion coarse screening result set, sending the characteristics into a second fine ranking model as supplementary data corpus, training, sorting and screening to obtain a diagnosis and treatment scheme and a diagnosis reminding service result set, and sending a recommendation result to a doctor end.

In a second aspect, an embodiment of the present invention provides a medical image guided diagnosis and guidance method based on artificial intelligence, including the following steps:

acquiring basic information, chief complaint information and AI inquiry exchange information of a patient;

generating a structured case according to the patient information, and pushing the case to a patient end and a doctor end respectively;

creating and integrating a chronic disease knowledge map, an image examination map and an image professional diagnosis data medical image data resource library;

constructing a matching model according to an integrated matching algorithm, a matching rule and a knowledge map retrieval rule, extracting features from a medical image data resource library, manufacturing an image data corpus, training the matching model by adopting a machine learning method, inputting case information into the matching model for matching, and obtaining a multi-path fusion coarse screening recall result set;

extracting features associated with patient services from the multi-path fusion coarse screening recall result set, sending the features as supplementary linguistic data into a first fine ranking model, training, sorting and screening to obtain diagnosis guide information, disease knowledge and risk prediction result sets, and sending a recommendation result to a patient end;

and extracting characteristics associated with the doctor service from the multi-path fusion coarse screening result set, sending the characteristics as supplementary data corpus into a second fine ranking model, training, sorting and screening to obtain a diagnosis and treatment scheme and a diagnosis reminding service result set, and sending a recommendation result to a doctor end.

In a third aspect, an intelligent device provided in an embodiment of the present invention includes a processor, an input device, an output device, and a memory, where the processor, the input device, the output device, and the memory are connected to each other, the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method described in the foregoing embodiment.

In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, in which a computer program is stored, the computer program including program instructions, which, when executed by a processor, cause the processor to execute the method described in the above embodiment.

The invention has the beneficial effects that:

the embodiment of the invention provides a medical image diagnosis guide inspection system, a method, equipment and a medium based on artificial intelligence, and particularly, the system fuses and uses computer artificial intelligence natural language processing technical solutions comprising machine learning, deep learning, entity identification, similarity matching, intention identification, a knowledge map, a recommendation system, an question-answering system, risk prediction and the like, and processes, trains and machine learning industrial data including professional image report real corpus data and image inspection professional academic data, so that auxiliary recommendation and question-answering consultation of artificial intelligence technologies such as professions, standardized inspection projects, diagnosis schemes and services aiming at medical images can be realized, and self-examination of patients and diagnosis of doctors are assisted.

Drawings

In order to more clearly illustrate the detailed description of the invention or the technical solutions in the prior art, the drawings that are needed in the detailed description of the invention or the prior art will be briefly described below. Throughout the drawings, like elements or portions are generally identified by like reference numerals. In the drawings, elements or portions are not necessarily drawn to scale.

Fig. 1 is a block diagram illustrating an artificial intelligence-based medical image guidance system according to a first embodiment of the present invention;

FIG. 2 is a flow chart of a method for guiding and examining medical images based on artificial intelligence according to a second embodiment of the present invention;

fig. 3 shows a block diagram of an intelligent device according to a third embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to a determination" or "in response to a detection". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".

It is to be noted that, unless otherwise specified, technical or scientific terms used herein shall have the ordinary meaning as understood by those skilled in the art to which the invention pertains.

As shown in fig. 1, a block diagram of a medical image guidance and inspection system based on artificial intelligence according to a first embodiment of the present invention is shown, the system includes: an inquiry module, a case generation module, a data resource module, a matching module, a diagnosis guide module and a detection guide module,

the inquiry module is used for collecting basic information, chief complaint information and AI inquiry exchange information of the patient;

the case generation module is used for generating structured cases from the patient information and pushing the cases to the patient end and the doctor end respectively;

the data resource module is used for creating and integrating medical image data resources such as a chronic disease knowledge map, an image examination map, image professional diagnosis data and the like;

the matching module is used for constructing a matching model according to an integrated matching algorithm, a matching rule and a knowledge map retrieval rule, extracting features from a medical image data resource library, manufacturing an image data corpus, training the matching model by adopting a machine learning method, inputting case information into the matching model for matching, and obtaining a multi-path fusion coarse screening recall (recommendation) result set;

the diagnosis guide module is used for extracting characteristics associated with the patient service from the multi-path fusion coarse screening result set, sending the characteristics as supplementary linguistic data into the first fine ranking model, training, sequencing and screening to obtain result sets such as popular science diagnosis guide information, disease knowledge, risk prediction and the like, and sending a recommendation result to a patient end;

the inspection guide module is used for extracting characteristics associated with doctor services from the multi-path fusion coarse screening result set, sending the characteristics into a second fine ranking model as supplementary data corpus, training, sorting and screening to obtain service result sets such as professional diagnosis and treatment schemes and diagnosis reminders, and sending recommendation results to a doctor end.

In this embodiment, the inquiry module includes a patient self-complaint unit, a recommended doctor unit, a big data inquiry and answer retrieval unit, an AI customer service inquiry and answer system unit, a doctor-patient communication unit, a recommended specialist answering unit, a patient information recording unit and an image data corpus unit; the patient self-complaint unit is used for collecting and acquiring personal information of a user, wherein the personal information comprises name, age, sex, region, symptom description, symptom attribute description, first complaint history information, fifth complaint history information, and the like, wherein the symptom attribute description comprises attack part, degree, frequency and the like; the recommending doctor unit is used for recommending associated doctors according to the patient complaint information, so that doctor-patient communication is facilitated; the big data question-answer searching unit is used for searching internet mainstream medical question-answer information and recommending experts to answer questions to patients; the AI customer service question-answering system unit is used for the on-line communication between the patient and the AI customer service, and consulting the question to obtain an answer; the patient information recording unit is used for recording patient self-complaint information, patient and AI customer service chat information, doctor-patient communication information and expert answering information and storing the information into a database; the image data corpus unit is used for taking out the recorded patient self-complaint information, the recorded patient and AI customer service chat information, the recorded doctor-patient communication information and the recorded expert answering information from the database, analyzing and inputting the image data corpus, returning the image data corpus to the AI customer service question-answering system unit for analysis and organization conversation, recycling the data and synchronously using the data for the training and learning of the subsequent matching module.

The AI customer service question-answering system unit comprises an entity identification unit, an intention identification unit and a map retrieval unit. The entity identification unit is used for carrying out strict retrieval matching and similarity approximate matching entity identification on the problems input by the patient, namely synonymy replacement, and generating standard expression problems; the intention recognition unit is used for sending the patient standard expression problems after matching or synonymous replacement processing into machine learning model training classification for problem classification, such as: sentences containing keywords of 'how to treat' and 'what medicine to take' tend to consult treatment method problems, and sentences containing keywords of 'how long to cure' and 'recovery' tend to consult treatment period problems; the atlas retrieval unit is used to guide patient questions to answer templates by question classification, such as: the treatment method with the answer template of the treatment method type questions being 'disease { } has { }', the answer template for inquiring the treatment period type questions is 'disease { } with the cure period being' and data resources such as a chronic disease knowledge map, an image knowledge map and the like are searched for keywords, the template is filled, if aiming at the keyword 'otitis media', the treatment method 'using ear drop phenol glycerol and oral anti-inflammatory drugs' is searched, the cure period is '10-14 days', complete answers are returned to the patient, and the treatment method for the disease otitis media has the following steps: the ear drops of phenol and glycerol and oral anti-inflammatory drugs are used, and the curing period of the otitis media with diseases is 10 to 14 days.

The case generation module comprises an information acquisition unit and a case generation unit, wherein the information acquisition module is used for reading data such as patient chief complaint information, and the case generation unit is used for generating patient information into structured case result data.

The data resource module comprises a chronic disease knowledge map unit, an image inspection map unit, an image report unit, a big data expert question-answering unit and a self-expansion knowledge map unit, wherein the chronic disease knowledge map unit is used for generating entity and relation pair map data from chronic disease department map knowledge data such as relational disease names, introduction, etiology, diseases, treatment, prevention, daily life and the like, and importing a chronic disease knowledge map database; the image examination map unit is used for generating map data of an entity and a relation pair from image professional data such as relational symptoms, examination, indications, diagnosis and treatment schemes, equipment, scanning modes, examination body positioning and the like, and importing the map data into an image examination knowledge map database; the image report unit is used for acquiring and acquiring an original image report data knowledge base; the big data expert question-answering unit is used for collecting Internet medical consultation expert question-answering data; the self-expansion knowledge map unit is used for extracting entity and relation pairs in patient cases, reports and Internet expert glume answering data, and automatically creating an expansion knowledge map in real time.

The matching module comprises a data processing unit, a map retrieval unit, a neural network model unit, a similarity feature extraction unit, a recommendation model unit and a recall result set unit.

The data processing unit is used for acquiring data in the chronic disease knowledge graph, the image inspection knowledge graph, the image report knowledge base, the big data expert for question answering and the self-expansion knowledge graph, analyzing, processing, extracting features, expressing the features by a mathematic vectorization symbol, manufacturing an image data corpus, and matching machine learning training intents.

The map retrieval unit is used for retrieving the chronic disease management knowledge map, the image examination knowledge map and the self-expansion knowledge map according to the patient case data, finding the correlation matching result entity relation pair and obtaining a map retrieval recall result set.

The neural network model unit is used for building a neural network model, sending image data corpora into the model, performing machine learning and training, performing intention analysis and prediction on patient case information, generating question and answer pairs and risk prediction candidate sets, obtaining a neural network result set, and assisting map retrieval and matching.

The similarity feature extraction unit is used for extracting similarity features from matching sentence pair data required by training a data making model in the image data corpus. The similarity feature extraction comprises the following steps: sentence length similarity extraction, longest public substring extraction, longest public subsequence extraction, edit distance feature extraction, n-gram feature word extraction, jieba (Chinese character of the Rong) participle feature word extraction, query word feature extraction, regular weight keyword feature extraction and word vector feature extraction. The sentence length similarity extraction is used for extracting the characteristics of the two sentence lengths of the matched sentences, and the similarity of intentions is measured by using the sentence length difference; the longest common substring extraction is used for extracting the longest identical character string length characteristics contained in two matched sentences, and the longest repeated words in the sentences are used for measuring the similarity; the longest public subsequence extraction is used for extracting the accumulated value characteristics of the longest identical character string length contained in the two matching parties, and the repeated word length in the sentence is used for measuring the similarity; the edit distance feature extraction is used for extracting the conversion complexity of the two matching parties, and the similarity features of the sentence length, the sentence sequence and the dimension of repeated words are measured by the conversion complexity of the two parties; the n-gram feature extraction is used for extracting and matching the features of the duplicate words of the two-character combination and the three-character combination of the two parties, and the similarity is measured by using the same number indexes of adjacent characters in the sentence; extracting the jieba word segmentation characteristics to extract the same number of the conventional words matched with the jieba word segmentation of the two parties, and measuring the similarity by using the repeatability indexes of the conventional words in the sentence; the query feature extraction is used for extracting and comparing the same query features of the two parties, positioning question types by using the queries and grasping the similarity of sentence bodies, such as: what is? "represents that the answer is inclined to an explanatory result, e.g., the introduction of the disease," how long? "represents that the answer is a time, e.g., recovery time; the rule weight keyword feature extraction is used for weighting the weight keywords contained by the matching parties, such as: both sides contain the knee, which indicates that the patient's current round of problem is to expand around the core of the knee, i.e. feature words related to the knee, the knee joint and other knee joints in the corpus can be weighted and reprocessed, and the score reference of similarity calculation is improved; the word vector feature extraction is used for performing mathematical vectorization symbolic processing on the sentences of the two matched parties and directly sending the sentences into a BilSTM (bidirectional cycle long-time memory) deep learning model and an Attention mechanism model (Bert) to learn similarity rules.

The recommendation model unit is used for sending the features extracted by the similarity feature extraction unit into the matching model, and learning the similarity features: the method comprises a collaborative filtering algorithm, matrix decomposition, a GBDT + LR (decision tree model + logistic regression) model, a Wide & Deep learning model and rule weight recall; a collaborative filtering algorithm for calculating a patient (UserCF) similarity based result set and a condition (ItemCF) similarity based result set; matrix factorization is used to decompose co-occurrence matrices in UserCF or ItemCF, computing associative similarities of patient and underlying semantics of condition information, such as: patient matrix, zhang san: [ chest distress 0.8 nausea 0.6 arrhythmia 0.9], symptom matrix, biliary heart disease: [ chest distress 0.81 nausea 0.61 arrhythmia 0.91], then the score of the Zhang Sandi Chongyuan heart disease is the probability of chest distress of the Zhang Sanchi heart disease + the frequency of the Zhang Sannawei heart disease + the probability of nausea of the Zhang Sannawei heart disease + the probability of heart rate irregularity of the Chongyuan heart disease, the part of the patient matrix and the symptom matrix which are commonly found is calculated and concentrated into an index, the similarity of the patient and the symptom is represented, the cooperative filtration only utilizes the interactive information of the patient and the symptom, the deficiency of the self attribute of the patient and the self attribute of the symptom is not used, the problem of sparse matrix is solved, and the generalization capability of the model is enhanced; the GBDT + LR (decision tree model + logistic regression) model is used for automatically screening and combining the characteristics of the context information by using a decision tree machine learning model to generate a new discrete characteristic vector, and generating a prediction result by using an LR (logistic regression) model to make up the defects that the cooperative filtering only uses the associated interactive information between the patient and the disease and neglects the characteristics of the patient's own characteristics and the disease information; the Wide & Deep learning model is used for roughly screening similarity by adopting a fast association rule of a Wide part of the Deep learning model, so that the direct memory capability of the model is enhanced, the Deep part of the Deep learning model goes Deep into the abstract training model, the abstract generalization capability of the model is enhanced, and the purpose of combining efficiency and accuracy is achieved; the rule weight recall is used for weighting and recalling the weight feature words in the similarity extraction features to obtain a rule weight result set, such as the problem related to knee, and carrying out weighted recalculation on similarity ranking on entry information related to knee in the image data corpus.

And the recall result set unit is used for splicing and normalizing multi-channel recall result sets such as a chart retrieval result set, a neural network result set, a matching model result set, a rule weight result set, a recommendation model result set and the like into dimensionless and standard quantization index probability values between 0 and 1 to obtain a multi-channel fusion coarse screening recall result set.

In this embodiment, the diagnosis guiding module includes a first refined ranking unit and a first recommendation application unit, and the first refined ranking unit includes a first supplementary service feature extraction unit and a first refined recommendation ranking model unit; the first supplementary service feature extraction unit is used for extracting features which are associated with the enhanced multi-path recall result set and used for the patient-side service, wherein the features include departments, examination items, disease knowledge, biographical guidance and risk prediction; and the first refined recommendation sequencing model unit is used for generating new data corpora by supplementing and extracting service features and sending the new data corpora into the model, and performing refined machine learning, training, similarity calculation, sequencing and screening to obtain a patient-side recommendation result set. The first recommendation application unit is used for combining patient-side recommendation results including departments, examination items, disease knowledge, biographical guidance and risk prediction with patient cases and recommending the patient-side recommendation results to the patient-side application.

In this embodiment, the doctor-side review module includes a recall result set second ranking unit and a second recommendation application unit, and the second ranking unit includes a second supplementary service feature extraction unit, a second ranking recommendation model unit and a key word reminding unit; the second supplementary service feature extraction unit is used for extracting features which are associated with services of a doctor end and enhanced from the multi-channel recall result set, wherein the features include professional examination items, item introduction, indications, equipment, scanning modes, examination body positioning and diagnosis and treatment schemes; the second refined recommendation sequencing model unit is used for sending new data corpora generated by the supplementary extraction service features into the model, and performing refined machine learning, training, similarity calculation, sequencing and screening to obtain a doctor-side recommendation result set; the key word reminding unit is used for highlighting and quickly reminding key words such as symptoms and risks in the patient case and identifying the symptoms and the risk key words in the patient case; the second recommending unit is used for recommending the doctor end recommending results including examination items, item introduction, indications, equipment, scanning modes, examination body positioning, diagnosis and treatment schemes, fast key word reminding and patient cases to the doctor end for application.

The embodiment of the invention provides a medical image diagnosis guide inspection system based on artificial intelligence, and particularly, the system integrates and uses a whole set of computer artificial intelligent Natural Language Processing (NLP) technical solutions comprising machine learning, deep learning, similarity matching, intention identification, knowledge maps, a recommendation system, an answering and questioning system, risk prediction and the like, and processes, trains and machine learning industrial data including professional, large-scale and systematized professional image report real corpus data and image inspection professional academic data, so that auxiliary recommendation and answering consultation of artificial intelligence technologies such as professions, standardized inspection projects, diagnosis schemes and services aiming at medical images can be realized, and self-examination and doctor diagnosis of patients are assisted.

In the first embodiment, a medical image guidance and inspection system based on artificial intelligence is provided, and correspondingly, another embodiment of the present application provides a medical image guidance and inspection method based on artificial intelligence, please refer to fig. 2, which shows a flowchart of a medical image guidance and inspection method based on artificial intelligence according to another embodiment of the present application. Since the method embodiment is basically similar to the device embodiment, the description is simple, and the relevant points can be referred to the partial description of the device embodiment. The method embodiments described below are merely illustrative.

As shown in fig. 2, a flowchart of a medical image guidance and detection method based on artificial intelligence according to another embodiment of the present invention is shown, and the method includes the following steps:

and S1, acquiring basic information, chief complaint information and AI inquiry exchange information of the patient.

Specifically, the information of chief complaints such as disease symptoms, medical history and personal information input by the patient in the chat window and other chat information with AI customer service are obtained, analyzed and stored.

The method specifically comprises the following steps:

collecting personal information of a user, wherein the personal information comprises name, age, sex, area, symptom description, symptom attribute description, first complaint and fifth complaint information, such as attack part, degree, frequency and the like, medical history, allergy history and the like;

recommending associated doctors according to the patient complaint information for doctor-patient communication;

retrieving internet mainstream medical question and answer information for recommending experts to answer questions;

and (4) the patient communicates with the AI customer service on line, consults the questions and obtains answers. The specific method comprises the following steps: entity identification, intention identification and map retrieval. In the entity identification step, the problems input by the patient are subjected to strict retrieval matching and similarity approximate matching entity identification, namely synonymy replacement, so that standard expression problems are generated; in the intention identification step, the patient standard expression problems after matching or synonymous replacement processing are sent to a machine learning model training classification for problem classification, such as: sentences containing keywords of 'how to treat' and 'what medicine to take' tend to consult treatment method problems, and sentences containing keywords of 'how long to cure' and 'recovery' tend to consult treatment period problems; the atlas retrieval step directs patient questions to answer templates by question classification, such as: the treatment method with the answer template of the treatment method type questions being 'disease { } has { }', the answer template for inquiring the treatment period type questions is 'disease { } with the cure period being' and data resources such as a chronic disease knowledge map, an image knowledge map and the like are searched for keywords, the template is filled, if aiming at the keyword 'otitis media', the treatment method 'using ear drop phenol glycerol and oral anti-inflammatory drugs' is searched, the cure period is '10-14 days', complete answers are returned to the patient, and the treatment method for the disease otitis media has the following steps: the ear drops of phenol and glycerol and oral anti-inflammatory drugs are used, and the curing period of the otitis media with diseases is 10 to 14 days.

Recording patient self-complaint information, chat information between the patient and AI customer service, doctor-patient communication information and expert answering information, and warehousing;

and taking out the recorded patient self-complaint information, the patient and AI customer service chat information, the doctor-patient communication information and the expert answering information from the database, analyzing and inputting an image data corpus, returning to an AI customer service question-answering system unit for analysis and organization conversation, and recycling data, and synchronously using the data for subsequent matching module training and learning.

And S2, generating a structured case according to the patient information, and pushing the case to the patient end and the doctor end respectively.

Specifically, data such as patient complaint information is acquired;

generating structured case result data from the patient information;

acquiring generated case result data and displaying information at a patient end;

and acquiring and generating case result data and making a case list and information display at a doctor end.

S3, creating and integrating medical image data resource library such as chronic disease knowledge map, image examination map, image professional diagnosis data, etc.

Specifically, the chart data of entity-relation pairs is generated by the chronic disease department chart knowledge data of relation type disease names, introduction, causes, diseases, treatment, prevention, daily life and the like, and is stored in a chronic disease knowledge chart database;

generating image data of entity and relation pairs by using image professional data such as relation diseases, examination, indications, diagnosis and treatment schemes, equipment, scanning modes, examination body positioning and the like, and storing the image data into an image examination knowledge map database;

acquiring an original image report data knowledge base;

acquiring Internet medical consultation expert answering data;

and extracting entity and relation pairs in patient cases, reports and Internet expert answering data, and automatically creating the extended knowledge graph in real time.

And S4, constructing a matching model according to the integrated matching algorithm, the matching rule and the knowledge map retrieval rule, extracting features from the medical image data resource library, making an image data corpus, training the matching model by adopting a machine learning method, inputting case information into the matching model for matching, and obtaining a multi-path fusion coarse screening recall (recommendation) result set.

Specifically, data in a chronic disease knowledge graph, an image inspection knowledge graph, an image report knowledge base, a big data expert question answering and a self-expansion knowledge graph are obtained, analyzed, processed, extracted, expressed by a mathematic vectorization symbol, and an image data corpus is manufactured and used for machine learning and training intention and matching. And searching the chronic disease management knowledge graph, the image examination knowledge graph and the self-expansion knowledge graph according to the patient case data, finding out a correlation matching result entity relationship pair, and obtaining a graph search recall result set. Building a neural network model, sending image data corpora into the model, performing machine learning and training, performing intention analysis and prediction on patient case information, generating question and answer pairs and risk prediction candidate sets, obtaining a neural network result set, and assisting map retrieval and matching. Matching sentences required by training a data making model in an image data corpus are used for data pair, and similarity features are extracted: extracting the characteristics of the sentence lengths of two parties of the matched sentences, and measuring the similarity of intentions by using the sentence length phase difference; extracting the longest identical character string length characteristics of both sides of the matched sentence, and measuring the similarity by using the longest repeated word in the sentence; extracting the accumulated value characteristics of the longest character string length contained in the two matched parties, and measuring the similarity by using the length of repeated words in the sentence; extracting the conversion complexity of the two matching parties, and measuring the similarity characteristics of the sentence length, the sentence sequence and the dimension of repeated words by using the conversion complexity of the two parties; extracting the characteristics of the duplicate words of the two-character combination and the three-character combination of the two matched parties, and measuring the similarity by using the same number indexes of adjacent characters in the sentence; extracting the same number of the conventional words after matching the jieba participles of the two parties, and measuring the similarity by using the repeatability indexes of the conventional words in the sentence; extracting and comparing the characteristics of the matched doubtful words, locating the problem types by the doubtful words and grasping the similarity of the sentence bodies; extracting a score reference for weighting the weight keywords contained by the two matched parties and improving similarity calculation; carrying out mathematical vectorization symbolic processing on the sentences of the two matched parties, and directly sending the sentences into a BilSTM (bidirectional cycle long-time memory) deep learning model and an Attention mechanism model (Bert) to learn similarity rules; sending the features extracted by the similarity feature extraction unit into a matching model, and learning similarity features: calculating a patient (UserCF) similarity based result set and a condition (ItemCF) similarity based result set; decomposing a co-occurrence matrix in UserCF or ItemCF, and calculating the relevance similarity of the implicit semantics of the patient and the disease information; the decision tree machine learning model is used for automatically carrying out feature screening combination on the context information to generate a new discrete feature vector, and a prediction result is generated through an LR (logistic regression) model, so that the defect that the cooperative filtering only utilizes the associated interaction information between the patient and the disease and ignores the self features of the patient and the information features of the disease is overcome; the similarity is roughly screened by adopting a rapid association rule of a Wide part of a Deep learning model, the direct memory capability of the model is enhanced, the Deep part goes Deep into an abstract training model, the abstract generalization capability of the model is enhanced, and the purpose of combining efficiency and accuracy is achieved; weighting and recalling the result of the weight feature words in the similarity extraction features; and splicing and normalizing the multi-channel recall result sets such as the atlas retrieval result set, the neural network result set, the matching model result set, the rule weight result set, the recommendation model result set and the like into dimensionless and standard quantization index probability values between 0 and 1 to obtain a multi-channel fusion coarse screening recall result set.

And S5, extracting features associated with the patient service from the multi-path fusion coarse screening result set, sending the features as supplementary data corpus into a first fine ranking model, training, sorting and screening to obtain result sets such as science popularization diagnosis information, disease knowledge and risk prediction, and sending a recommendation result to a patient side.

Specifically, extracting features which are associated with the enhanced patient-side business from the multi-recall result set and serve as first extracted supplementary business features, wherein the first extracted supplementary business features comprise departments, examination items, disease knowledge, biographical guidance and risk prediction;

generating a new data corpus by the first supplementary extraction service characteristics, sending the new data corpus into a first refined model, and performing refined machine learning, training, similarity calculation, sequencing and screening to obtain a patient-side science popularization recommendation result set;

and recommending results including departments, examination items, disease knowledge, biographical guidance, risk prediction and patient cases to the patient end, and recommending the results to the patient end for application.

And S6, extracting the characteristics associated with the doctor service from the multi-path fusion coarse screening result set, sending the characteristics as supplementary data corpus into a second fine ranking model, training, sorting and screening to obtain a professional diagnosis and treatment scheme, a diagnosis prompt and other service result sets, and sending a recommendation result to a doctor end.

Specifically, the characteristics of the multi-channel recall result set, which are extracted and strengthened to be associated with the doctor-side service, are used as second extracted supplementary service characteristics, including professional examination items, item introduction, indications, equipment, scanning modes, examination body positioning and diagnosis and treatment schemes;

generating a new data corpus by using the second supplementary extraction service characteristics, sending the new data corpus into a second refined model, and performing refined machine learning, training, similarity calculation, sequencing and screening to obtain a professional doctor-side recommendation result set;

highlighting key words such as symptoms, risks and the like in a patient case for prompt and prompt;

and recommending results to the doctor end, wherein the results comprise examination items, item introduction, indications, equipment, scanning modes, examination body positioning, diagnosis and treatment schemes, and the combination of fast key word reminding and patient cases, and are recommended to the doctor end for application.

The above is a description of an embodiment of a medical image guidance and inspection method based on artificial intelligence according to a second embodiment of the present invention.

The medical image diagnosis and guidance method based on artificial intelligence and the medical image diagnosis and guidance system based on artificial intelligence are based on the same inventive concept, have the same beneficial effects, and are not repeated herein.

As shown in fig. 3, a block diagram of an intelligent device according to an embodiment of the present invention is shown, where the intelligent device includes a processor, an input device, an output device, and a memory, where the processor, the input device, the output device, and the memory are connected to each other, the memory is used for storing a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method described in the second embodiment.

It should be understood that in the embodiments of the present invention, the Processor may be a Central Processing Unit (CPU), and the Processor may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The input device may include a touch pad, a fingerprint sensor (for collecting fingerprint information of a user and direction information of the fingerprint), a microphone, etc., and the output device may include a display (LCD, etc.), a speaker, etc.

The memory may include both read-only memory and random access memory, and provides instructions and data to the processor. The portion of memory may also include non-volatile random access memory. For example, the memory may also store device type information.

In a specific implementation, the processor, the input device, and the output device described in the embodiments of the present invention may execute the implementation described in the method embodiments provided in the embodiments of the present invention, and may also execute the implementation described in the system embodiments in the embodiments of the present invention, which is not described herein again.

An embodiment of a computer-readable storage medium is also provided in the present invention, the computer-readable storage medium storing a computer program comprising program instructions that, when executed by a processor, cause the processor to perform the method described in the second embodiment above.

The computer readable storage medium may be an internal storage unit of the terminal described in the foregoing embodiment, for example, a hard disk or a memory of the terminal. The computer readable storage medium may also be an external storage device of the terminal, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like provided on the terminal. Further, the computer-readable storage medium may also include both an internal storage unit and an external storage device of the terminal. The computer-readable storage medium is used for storing the computer program and other programs and data required by the terminal. The computer readable storage medium may also be used to temporarily store data that has been output or is to be output.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the terminal and the unit described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed terminal and method can be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may also be an electric, mechanical or other form of connection.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the present invention, and they should be construed as being included in the following claims and description.

Claims (10)

Translated fromChinese

1.一种基于人工智能的医学影像导诊导检系统，其特征在于，包括：问诊模块、病例生成模块、数据资源模块、匹配模块、导诊模块和导检模块，1. a medical image guidance system based on artificial intelligence, is characterized in that, comprises: interrogation module, case generation module, data resource module, matching module, guiding module and guiding module,所述问诊模块用于收集患者基本信息、主诉信息及AI问诊交流信息；The interrogation module is used to collect basic patient information, chief complaint information and AI interrogation and communication information;所述病例生成模块用于将患者信息生成结构化病例，将病例分别推送给患者端和医生端；The case generation module is used for generating structured cases from patient information, and pushing the cases to the patient side and the doctor side respectively;所述数据资源模块用于创建、整合慢病知识图谱、影像检查图谱、影像专业诊断资料医学影像数据资源库；The data resource module is used to create and integrate a chronic disease knowledge map, an image examination map, and a medical image data repository of professional imaging diagnostic data;所述匹配模块用于根据整合匹配算法、匹配规则、知识图谱检索规则构建匹配模型，从医学影像数据资源中提取特征，制作影像数据语料库，采用机器学习方法训练匹配模型，将病例信息输入匹配模型进行匹配，得到多路融合粗筛召回结果集；The matching module is used to construct a matching model according to an integrated matching algorithm, matching rules, and knowledge map retrieval rules, extract features from medical image data resources, create a corpus of image data, use machine learning methods to train the matching model, and input case information into the matching model. Matching is performed to obtain a multi-channel fusion coarse screening recall result set;所述导诊模块用于将多路融合粗筛召回结果集提取与患者业务相关联特征，作为补充语料送入第一精排模型，进行训练、排序、筛选，得到导诊信息、疾病知识和风险预测结果集，向患者端发送推荐结果；The guide module is used to extract the features associated with the patient's business from the multi-channel fusion coarse screening recall result set, and send it into the first fine ranking model as a supplementary corpus for training, sorting, and screening to obtain guide information, disease knowledge and information. Risk prediction result set, sending recommended results to the patient;所述导检模块用于将多路融合粗筛召回结果集提取与医生业务相关联特征，作为补充数据语料送入第二精排模型，进行训练、排序、筛选，得到诊疗方案和诊断提醒服务结果集，向医生端发送推荐结果。The guided inspection module is used to extract the features associated with the doctor's business from the multi-channel fusion coarse screening recall result set, and send it into the second fine sorting model as a supplementary data corpus for training, sorting, and screening, and obtaining a diagnosis and treatment plan and a diagnosis reminder service. Result set, send the recommended results to the doctor.2.如权利要求1所述的系统，其特征在于，所述匹配模块包括数据处理单元和图谱检索单元，所述数据处理单元获取慢病知识图谱、影像检查知识图谱、影像报告知识库、大数据专家答疑、自扩展知识图谱中的数据，分析处理，提取特征，用数学向量化符号表示，制作影像数据语料库，用于机器学习训练意图与匹配；所述图谱检索单元用于根据患者病例数据检索慢病管理知识图谱、影像检查知识图谱和自扩展知识图谱，找到关联匹配结果实体关系数据，得到图谱检索召回结果集。2. The system according to claim 1, wherein the matching module comprises a data processing unit and a graph retrieval unit, and the data processing unit obtains a chronic disease knowledge graph, an image inspection knowledge graph, an image report knowledge base, a large Data experts answer questions, analyze and process the data in the self-expanding knowledge map, extract features, express it with mathematical vectorized symbols, and create a corpus of image data for machine learning training intent and matching; the map retrieval unit is used for data based on patient cases. Retrieve the chronic disease management knowledge map, image inspection knowledge map and self-expanding knowledge map, find the entity relationship data of the associated matching results, and obtain the map retrieval recall result set.3.如权利要求2所述的系统，其特征在于，所述匹配模块还包括神经网络模型单元、相似度特征提取单元、推荐模型单元和召回结果集单元，所述神经网络模型单元用于构建神经网络模型，将医学影像数据送入神经网络模型进行训练，对患者病例信息做意图分析和预测，生成神经网络结果集；3. The system of claim 2, wherein the matching module further comprises a neural network model unit, a similarity feature extraction unit, a recommendation model unit and a recall result set unit, and the neural network model unit is used to construct Neural network model, which sends medical image data into the neural network model for training, analyzes and predicts the patient's case information, and generates a neural network result set;所述相似度特征提取单元用于将影像数据语料库中的数据制作成神经网络模型训练所需的匹配句子对数据，提取相似性特征；The similarity feature extraction unit is used for making the data in the image data corpus into matching sentence pair data required for neural network model training, and extracting similarity features;所述推荐模型单元用于将相似性特征送入匹配模型中学习相似性特征，得到与患者主诉信息相似性结果集，推荐结果集；The recommendation model unit is used to send the similarity feature into the matching model to learn the similarity feature, and obtain a similarity result set with the patient's chief complaint information, and a recommendation result set;所述召回结果集单元用于将得到的多路召回结果集拼接，归一化为0到1之间的去量纲、标量化指标概率值，得到多路融合粗筛召回结果集。The recall result set unit is used for splicing the obtained multi-channel recall result set, normalizing it into a dimensionless and scalar index probability value between 0 and 1, and obtaining a multi-channel fusion coarse screening recall result set.4.如权利要求3所述的系统，其特征在于，所述导诊模块包括第一精排单元和第一推荐应用单元，4. The system of claim 3, wherein the guide module comprises a first fine-arrangement unit and a first recommended application unit,所述第一精排单元用于将多路召回结果集提取与患者相关联业务特征，将提取的与患者相关联业务特征生成新的数据语料送入第一精排模型进行训练、排序和筛选，得到导诊信息、疾病知识和风险预测结果集，得到患者端推荐结果集；The first fine sorting unit is used to extract the business features associated with the patient from the multi-way recall result set, and generate a new data corpus from the extracted business features associated with the patient and send it to the first fine sorting model for training, sorting and screening. , get the guide information, disease knowledge and risk prediction result set, and get the patient-side recommendation result set;所述第一推荐应用单元用于根据患者端推荐结果集向患者端推荐应用；The first recommended application unit is used for recommending applications to the patient side according to the patient side recommendation result set;所述导检模块包括第二精排单元和第二推荐应用单元，The guided inspection module includes a second fine-arrangement unit and a second recommended application unit,所述第二精排单元用于将多路召回结果集提取与医生相关联业务特征，将提取的与医生相关联业务特征生成新的数据语料送入第二精排模型进行进行训练、排序和筛选，得到诊疗方案和诊断提醒服务结果集，得到医生端推荐结果集，标识出患者病例中的病症和风险重点词；The second fine sorting unit is used to extract the business features associated with doctors from the multi-way recall result set, and generate new data corpus from the extracted business features associated with doctors and send them to the second fine sorting model for training, sorting and sorting. Screen, get the result set of diagnosis and treatment plan and diagnosis reminder service, get the result set recommended by the doctor, and identify the symptoms and risk key words in the patient's case;所述第二推荐应用单元用于根据医生端推荐结果集向医生端推荐应用。The second recommendation application unit is used for recommending applications to the doctor side according to the doctor side recommendation result set.5.如权利要求1所述的系统，其特征在于，所述问诊模块包括AI客服问答系统单元，所述AI客服问答系统单元包括实体识别单元、意图识别单元和图谱检索单元；5. The system according to claim 1, wherein the inquiry module comprises an AI customer service question and answer system unit, and the AI customer service question and answer system unit comprises an entity recognition unit, an intention recognition unit and a map retrieval unit;所述实体识别单元用于将患者输入的问题进行检索匹配和相似度近似匹配实体识别，生成标准表述问题；The entity recognition unit is used to perform retrieval matching and similarity approximate matching entity recognition on the problem input by the patient, and generate a standard expression problem;所述意图识别单元用于将标准表述问题送入机器学习模型训练分类，进行问题分类；The intention recognition unit is used for sending the standard expression problem into the machine learning model for training classification, and classifying the problem;所述图谱检索单元用于将患者问题按问题分类导向答案模板。The atlas retrieval unit is used to guide patient questions to answer templates by question classification.6.一种基于人工智能的医学影像导诊导检方法，其特征在于，包括以下步骤：6. An artificial intelligence-based medical image guide diagnosis and inspection method, characterized in that, comprising the following steps:获取患者基本信息、主诉信息及AI问诊交流信息；Obtain basic patient information, chief complaint information and AI consultation and communication information;根据患者信息生成结构化病例，将病例分别推送给患者端和医生端；Generate structured cases based on patient information, and push the cases to the patient side and the doctor side respectively;创建、整合慢病知识图谱、影像检查图谱、影像专业诊断资料医学影像数据资源库；Create and integrate chronic disease knowledge maps, imaging inspection maps, and medical imaging data resources for professional imaging diagnostic data;根据整合匹配算法、匹配规则、知识图谱检索规则构建匹配模型，从医学影像数据资源库中提取特征，制作影像数据语料库，采用机器学习方法训练匹配模型，将病例信息输入匹配模型进行匹配，得到多路融合粗筛召回结果集；Build a matching model based on the integrated matching algorithm, matching rules, and knowledge map retrieval rules, extract features from the medical image data repository, create an image data corpus, use machine learning methods to train the matching model, and input the case information into the matching model for matching. Road fusion coarse screening recall result set;将多路融合粗筛召回结果集提取与患者业务相关联特征，作为补充语料送入第一精排模型，进行训练、排序和筛选，得到导诊信息、疾病知识和风险预测结果集，向患者端发送推荐结果；The multi-channel fusion coarse screening recall result set is extracted and the features associated with the patient's business are sent to the first fine-arrangement model as a supplementary corpus for training, sorting and screening to obtain the guide information, disease knowledge and risk prediction result set. The terminal sends the recommendation result;将多路融合粗筛召回结果集提取与医生业务相关联特征，作为补充数据语料送入第二精排模型，进行训练、排序和筛选，得到诊疗方案和诊断提醒服务结果集，向医生端发送推荐结果。The multi-channel fusion coarse screening recall result set is extracted and the features associated with the doctor's business are sent to the second refined ranking model as a supplementary data corpus for training, sorting and screening to obtain the diagnosis and treatment plan and diagnosis reminder service result set, and send it to the doctor. Recommended results.7.如权利要求6所述的方法，其特征在于，所述将慢病知识图谱、影像检查图谱、影像专业诊断资料整合成医学影像数据资源具体包括：7. The method of claim 6, wherein the integration of chronic disease knowledge maps, imaging examination maps, and imaging professional diagnostic data into medical image data resources specifically includes:将关系型病名、介绍、病因、病症、治疗、预防和日常慢病科普知识数据，生成实体与关系对的图数据，存入慢病知识图谱数据库；The relational disease name, introduction, etiology, symptoms, treatment, prevention, and daily chronic disease popularization knowledge data are generated to generate entity-relationship graph data, and stored in the chronic disease knowledge graph database;将关系型病症、检查、适应症、诊疗方案、设备、扫描方式和检查身体摆位影像专业资料数据，生成实体与关系对的图数据，存入影像检查知识图谱图数据库；The relational diseases, examinations, indications, diagnosis and treatment plans, equipment, scanning methods, and professional image data of examination body placement images are generated, and the graph data of entity and relation pairs are generated, and stored in the image examination knowledge graph database;采集获取原始影像报告数据知识库；Collect and obtain the original image report data knowledge base;采集互联网医疗咨询专家答疑数据；Collect the data of Internet medical consultation experts answering questions;提取患者病例、报告、互联网专家答疑数据中的实体与关系对，实时自动创建扩展知识知识图谱。Extract entity-relationship pairs from patient cases, reports, and Internet expert Q&A data, and automatically create extended knowledge knowledge graphs in real time.8.如权利要求7所述的方法，其特征在于，所述从医学影像数据资源中提取特征，制作影像数据语料库，采用机器学习方法训练匹配模型，将病例信息输入匹配模型进行匹配，得到多路融合粗筛召回结果集具体包括：8. The method according to claim 7, wherein the feature is extracted from medical image data resources, an image data corpus is made, a machine learning method is used to train a matching model, and the case information is input into the matching model for matching, and multiple data are obtained. The road fusion coarse screening recall result set specifically includes:获取慢病知识图谱、影像检查知识图谱、影像报告知识库、大数据专家答疑、自扩展知识图谱中的数据，分析处理，提取特征，用数学向量化符号表示，制作影像数据语料库，用于机器学习训练意图与匹配；Acquire data from chronic disease knowledge graph, image inspection knowledge graph, image report knowledge base, big data expert Q&A, and self-expanding knowledge graph, analyze and process, extract features, represent them with mathematical vectorized symbols, and make image data corpus for machine use Learning training intent and matching;根据患者病例数据检索慢病管理知识图谱、影像检查知识图谱、自扩展知识图谱，找到关联匹配结果实体关系对，得到图谱检索召回结果集；Retrieve chronic disease management knowledge map, image inspection knowledge map, and self-expanding knowledge map according to patient case data, find the entity-relationship pair of associated matching results, and obtain a map retrieval recall result set;搭建神经网络模型，将影像数据语料送入神经网络模型，机器学习、训练，对患者病例信息作意图分析、预测，用于生成问答对、风险预测候选集，辅助图谱检索与匹配，得到神经网络结果集；Build a neural network model, feed the image data corpus into the neural network model, perform machine learning and training, conduct intention analysis and prediction on patient case information, and use it to generate question-and-answer pairs and risk prediction candidate sets, assist map retrieval and matching, and obtain neural network. result set;将影像数据语料库中的数据制作成神经网络模型训练所需的匹配句子对数据，提取相似性特征；The data in the image data corpus is made into matching sentence pair data required for neural network model training, and similarity features are extracted;将相似性特征送入匹配模型中学习相似性特征，得到与患者主诉信息相似性结果集，推荐结果集；The similarity feature is sent into the matching model to learn the similarity feature, and the similarity result set with the patient's main complaint information is obtained, and the result set is recommended;将得到的多路召回结果集拼接，归一化为0到1之间的去量纲、标量化指标概率值，得到多路融合粗筛召回结果集。The obtained multi-channel recall result set is spliced, and normalized into a dimensionless and scalar index probability value between 0 and 1, and a multi-channel fusion coarse screening recall result set is obtained.9.一种智能设备，包括处理器、输入设备、输出设备和存储器，所述处理器、输入设备、输出设备和存储器相互连接，所述存储器用于存储计算机程序，所述计算机程序包括程序指令，其特征在于，所述处理器被配置用于调用所述程序指令，执行如权利要求6-8任一项所述的方法。9. A smart device comprising a processor, an input device, an output device and a memory, the processor, the input device, the output device and the memory being interconnected, the memory being used to store a computer program, the computer program comprising program instructions , characterized in that the processor is configured to invoke the program instructions to execute the method according to any one of claims 6-8.10.一种计算机可读存储介质，其特征在于，所述计算机存储介质存储有计算机程序，所述计算机程序包括程序指令，所述程序指令当被处理器执行时使所述处理器执行如权利要求6-8任一项所述的方法。10. A computer-readable storage medium, characterized in that the computer storage medium stores a computer program, the computer program comprising program instructions, the program instructions, when executed by a processor, cause the processor to execute as claimed The method of any of claims 6-8.

Images