CN117352159A - Method, system and storage medium for evidence-based treatment of difficult and complicated diseases based on electronic medical records - Google Patents
Method, system and storage medium for evidence-based treatment of difficult and complicated diseases based on electronic medical records Download PDFInfo
- Publication number
- CN117352159A CN117352159A CN202311276855.2A CN202311276855A CN117352159A CN 117352159 A CN117352159 A CN 117352159A CN 202311276855 A CN202311276855 A CN 202311276855A CN 117352159 A CN117352159 A CN 117352159A
- Authority
- CN
- China
- Prior art keywords
- electronic medical
- agent
- medical record
- diagnosis
- evidence
- 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
- 201000010099 disease Diseases 0.000 title claims abstract description 89
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 89
- 238000000034 method Methods 0.000 title claims abstract description 88
- 238000011282 treatment Methods 0.000 title claims abstract description 17
- 238000003860 storage Methods 0.000 title claims abstract description 13
- 238000003745 diagnosis Methods 0.000 claims abstract description 80
- 238000003748 differential diagnosis Methods 0.000 claims abstract description 60
- 238000012216 screening Methods 0.000 claims abstract description 23
- 230000008520 organization Effects 0.000 claims abstract description 16
- 230000002452 interceptive effect Effects 0.000 claims abstract description 9
- 230000008485 antagonism Effects 0.000 claims abstract description 4
- 239000003795 chemical substances by application Substances 0.000 claims description 190
- 230000008569 process Effects 0.000 claims description 34
- 230000003993 interaction Effects 0.000 claims description 23
- 229940039227 diagnostic agent Drugs 0.000 claims description 17
- 239000000032 diagnostic agent Substances 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 16
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 8
- 238000004088 simulation Methods 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 238000013138 pruning Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 5
- 238000005284 basis set Methods 0.000 claims description 3
- 230000006399 behavior Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 230000000977 initiatory effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000002457 bidirectional effect Effects 0.000 claims description 2
- 238000010845 search algorithm Methods 0.000 abstract 1
- 208000004987 Macrophage activation syndrome Diseases 0.000 description 21
- 201000001853 McCune-Albright syndrome Diseases 0.000 description 21
- 201000005085 Meconium Aspiration Syndrome Diseases 0.000 description 21
- 238000007689 inspection Methods 0.000 description 9
- 238000012545 processing Methods 0.000 description 9
- 208000024891 symptom Diseases 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 239000008188 pellet Substances 0.000 description 3
- 238000000342 Monte Carlo simulation Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 239000008186 active pharmaceutical agent Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000007717 exclusion Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 239000003607 modifier Substances 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 208000019901 Anxiety disease Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000036506 anxiety Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000013399 early diagnosis Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008506 pathogenesis Effects 0.000 description 1
- 238000002948 stochastic simulation Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
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
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- 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
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Epidemiology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a method, a system and a storage medium for evidence-based treatment of problematic diseases based on electronic medical records, which comprises the steps of firstly constructing a multi-agent differential diagnosis model framework, then screening similar associated historical electronic medical records from a historical electronic medical record database and a knowledge base according to patient electronic medical records, acquiring a differential discussion UCT tree of a current differential gist by using a UCT search algorithm in the screened electronic medical records by a differential agent, and carrying out structural organization and representation of the differential discussion; and finally, making the implicit authentication and arguments UCT tree constructed by each agent dominant to generate an authentication and arguments game forest of a plurality of agent groups, and simultaneously realizing interactive combined antagonism learning through zero and game of a plurality of agents to generate a common diagnosis and diagnosis arguments of the problematic diseases containing positive evidence and negative evidence. The invention has autonomy and interpretability of differential diagnosis evidence-based, can provide positive evidence for diagnosing diseases and negative evidence for eliminating diseases, and assists doctors in improving the accuracy of differential diagnosis.
Description
Technical Field
The invention relates to the technical field of medical artificial intelligence, in particular to a method, a system and a storage medium for evidence-based treatment of a problematic disease based on an electronic medical record.
Background
"problematic disease" refers to a disease with unknown etiology, atypical symptoms, unclear pathogenesis, and difficult diagnosis by repeated diagnosis; high mortality rate and is extremely easy to cause misdiagnosis and missed diagnosis. Bringing great burden to the body, spirit and economy of the patient. Studies have shown that delayed diagnosis and treatment of problematic disease results in an average of $ 2 tens of thousands of dollars per patient; and patients have lower quality of life and higher incidence of depression and anxiety. Finding a high-efficiency accurate method for diagnosing difficult and complicated diseases has important significance for early diagnosis of patients, increasing survival rate and improving diagnosis experience.
The differential diagnosis (Differential Diagnosis) is a dialectical analysis process of doctors around the contradiction and suspicious points of the diagnosis of the difficult and complicated diseases, and is used for eliminating the interference of similar diseases, obtaining the correct diagnosis conclusion and avoiding misdiagnosis and missed diagnosis. The comprehensive disease judgment method mainly refers to the medical history of a patient, multi-system physical examination and laboratory examination, is characterized in that the comprehensive disease judgment method is used for identifying other similar diseases in symptoms, signs and disease types through multiple disciplines and multiple visual angles, and is verified through a strict evidence-based elimination method, the differential diagnosis plays an important role in improving the diagnosis accuracy of the difficult and complicated diseases, but has high requirements on the medical knowledge, experience breadth and depth of doctors. There is a need for an intelligent solution that can automatically generate a differential diagnosis basis from patient medical records and historical cases.
Disclosure of Invention
In view of the above-mentioned shortcomings in the prior art, the present invention aims to provide a method, a system and a storage medium for evidence-based treatment of problematic diseases based on electronic medical records, which can search, organize and represent typical symptoms and similar disease symptoms related to the diseases of patients from the electronic medical records, and provide interpretable positive support evidence and negative exclusion evidence for intelligently and accurately judging the disease types and conditions of the patients.
In a first aspect, a method for evidence-based treatment of a problematic disease based on an electronic medical record is provided, comprising:
s1: establishing a multi-agent differential diagnosis model framework, constructing a plurality of agent models including diagnosis agents and differential agents, and representing a differential-diagnostic countermeasure process model by using zero and game models of the plurality of agents;
s2: acquiring an electronic medical record of a patient, and searching and screening similar associated historical electronic medical records in a historical electronic medical record database and a knowledge base by a diagnosis agent;
s3: the authentication agent adopts UCT searching algorithm to obtain and evaluate authentication argument UCT tree of the current authentication key point in the screened similar association history electronic medical record;
s4: according to the authentication argument UCT tree, carrying out structural organization and representation of the authentication argument for the argument interaction of multiple intelligent agents;
S5: the implicit identification theory data UCT tree constructed by each agent is made explicit to generate an identification theory game forest of a multi-agent group, thereby realizing theory element organization in the same state space, and simultaneously realizing interactive combined antagonism learning through zero and game of a plurality of agents to generate a difficult disease identification diagnosis common theory data set containing positive evidence and negative evidence.
Further, the step S1 includes:
the diagnosis agent and the identification agent are agent models comprising a sensing unit, an reasoning unit and an interaction unit; the diagnosis agent is responsible for controlling the initialization of differential diagnosis, the initiation of the identification issue of the problematic disease, the management, the control and the result judgment of the identification evidence-based process of the problematic disease; the identification agent is an reasoning search evidence-based executor for identifying the evidence-based process, and executes the discussion data construction behavior according to the evidence-based link where the evidence-based executor is located;
the discrimination-diagnostic countermeasure process model is represented in a zero and game model of a plurality of agents, which is represented as a talking dialogue game of a diagnostic agent and a plurality of discriminating agents: firstly, a diagnosis focus theory is provided by a diagnosis agent, then the diagnosis agent and a plurality of identification agents continuously provide attack theory around the focus theory to attack theory provided by the opposite party, and an allowable set containing the diagnosis focus theory is constructed through theory interaction so as to prove that the diagnosis focus theory is acceptable or can be abolished.
Further, the step S2 includes:
s21: extracting the characteristics of the electronic medical record of the patient, and the electronic medical records of the history electronic medical record database and the knowledge base to obtain the distribution probability of each subject term of the electronic medical record, and selecting N subject terms with the highest probability as the document subjects of the corresponding electronic medical record, wherein N is a preset value;
s22: the method comprises the steps of searching associated historical medical records, firstly, scanning a historical electronic medical record database and a knowledge base once based on a document theme of the patient electronic medical record, and counting the occurrence times of corresponding subject words in each historical electronic medical record to form a first candidate set; screening a first frequent item set with association rules meeting minimum strength according to a minimum support threshold; combining the first frequent item sets to form a second candidate item set; performing second scanning on the historical electronic medical record database and the knowledge base, counting each second candidate item set according to the first candidate item set counting method, and screening out second frequent item sets according to the minimum support threshold; iterative screening is carried out according to the scanning and screening method until the candidate item set is empty; and according to the finally generated frequent item set, generating the association between the electronic medical record of the patient and the historical electronic medical record by calculating the similarity between the historical electronic medical record and the electronic medical record of the patient in the finally generated frequent item set, removing the electronic medical record with the similarity lower than a similarity threshold value, and obtaining the candidate set of the identification agent search and the generation of the identification diagnosis evidence.
Further, the step S21 includes:
identifying and extracting entities and relations in the electronic medical record by using a trained bidirectional encoder model based on a transducer;
and obtaining the distribution probability of each entity or relation subject word of the electronic medical record by using the document subject generation model, and selecting N subject words with the highest probability as the document subjects of the corresponding electronic medical record.
Further, the step S3 includes:
s31: authentication arguments UCT tree initialization: taking the identification key points of the problematic diseases as the root nodes of the UCT tree in the initial state, wherein the root nodes store the electronic medical record data of the current patient;
s32: discrimination argues about the selection and expansion of the UCT tree: calculating the confidence upper bound of each child node according to the UCB algorithm, and selecting the child node with the maximum confidence upper bound for expansion so as to update the current authentication argument UCT tree; each child node stores a similar association history electronic medical record obtained by screening;
s33: discrimination arguments are the simulation and backtracking of the UCT tree: calculating the relation between the characteristics of the historical electronic medical record and the characteristics of patient complaints, current medical history, abnormal signs and checking indexes in the current electronic medical record of the patient, updating the access times, accumulated return values and average return values of each node of the whole authentication and discussion UCT tree from the child node to the root node, and pruning the authentication and discussion UCT tree according to the return estimation values to form a new authentication and discussion UCT tree, wherein the authentication and discussion UCT tree is formed by the root node of the current electronic case of the patient, the pruned historical electronic medical record child node and the supporting relation return estimation value among the nodes;
S34: and repeating the steps S32 and S33 until the evaluation value of the authentication and discussion data UCT tree converges or the preset termination condition of the upper limit of the searching time or the upper limit of the times is reached, and obtaining the final authentication and discussion data UCT tree, namely storing the optimal authentication and discussion data set of the historical electronic medical record.
Further, in the step S4, the identification argument elements are organized by using Toulmin argument six-tuple, so as to construct a complete identification argument representation; the discriminant theory structure Arg-Evi is represented as follows:
Arg-Evi＝{Diagnosis，Differential，Warrant，Backing，Strength，Data}
wherein diagnostis represents the proposition of a focus of Diagnosis of problematic diseases and similar diseases of diagnostic agents; the Differential represents an identification agent identification evidence element, which is the characteristic extracted from the patient electronic medical record stored in the identification data UCT tree; warrant represents the mapping relation of electronic medical record data to authentication evidence elements of an authentication agent; backsight represents identifying supplementary evidence elements stored by sub-nodes in the UCT tree as secondary arguments; strength represents the intensity estimate of the Differential; data represents raw Data in the historic electronic medical record and the current patient electronic medical record.
Further, the step S5 includes:
after constructing the authentication argument UCT tree, the implicit authentication argument of each agent is made dominant, and the implicit authentication argument is expressed as an element for differential diagnosis in an EAS empirical proof mode;
Adopting an Arena joint learning model in zero and game to perform joint learning, realizing interaction and comparison of arguments among agents, taking the winning arguments after multiple rounds of game as optimal global arguments for the current identification key points, which are superior to other agents, achieving a joint arguments set through calculation, and extracting local arguments to be stored as group arguments;
the multi-agent common arguments are used as a common diagnosis and differential diagnosis decision basis set to carry out spiral evolution of the arguments game tree, and a multi-agent arguments game forest is generated to be used as the differential diagnosis basis.
Further, the Arena joint learning model is as follows:
wherein, disease is the issue of differential diagnosis, dia is the diagnostic agent, dis-Dia is the set of differential agents, arg is the focus of differential diagnosis disputes, UCT-Tree is the set of differential data UCT Tree constructed by the agents,for the attack theory data set of structural organization and representation, MAS-AGF is a differential diagnosis multi-agent theory game framework, GK is a high-quality global knowledge set which is jointly and interactively learned by the multi-agent aiming at the differential diagnosis dispute focus Arg in the current subject treatment;
the multi-agent joint theory data set construction process combined with Arena joint learning model is as follows: the diagnosis agent broadcasts to the identification agent set according to the events and Arg input by the user, the identification agent set realizes role conversion and multiparty discussion game according to MAS-AGF to eliminate inconsistent evidence in UCT-Tree, In particular byCarrying out conflict resolution and consistency fusion on the estimated value inconsistent evidence to obtain a consensus-achieved claim and a diagnosis conclusion supported by the consensus-achieved claim; the final winning discrimination agent comprises ++>The method is a high-quality global knowledge obtained by multi-agent joint learning;
and repeatedly completing the joint learning process by all the identification intelligent agents to obtain a high-quality global knowledge set GK related to the current patient case event as a common argumentation set, namely, the identification diagnosis evidence of the possible diseases, which is searched and generated in a historical electronic medical record library and a knowledge base by the multi-intelligent agent aiming at the current patient case event.
In a second aspect, there is provided a system for evidence-based treatment of a difficult disease, comprising:
a memory having a computer program stored thereon;
and the processor is used for realizing the evidence-based method for the problematic diseases based on the electronic medical record according to any one of the above when the computer program is loaded and executed.
In a third aspect, a computer readable storage medium is provided, on which a computer program is stored, which when executed by a processor implements the method for evidence-based on electronic medical records of the problematic disease as described in any of the above.
The invention provides a method, a system and a storage medium for evidence-based treatment of problematic diseases based on electronic medical records, which are characterized in that firstly, a multi-agent differential diagnosis model framework is constructed, then similar associated historical electronic medical records are retrieved and screened from a historical electronic medical record database and a knowledge base according to patient electronic medical records, each differential agent adopts a UCT searching algorithm to acquire and evaluate differential arguments UCT tree of the current differential gist in the screened similar associated historical electronic medical records, and the differential arguments UCT tree is used for the structural organization and representation of the differential arguments for the interaction of the multi-agents; and finally, making the implicit authentication and arguments UCT tree constructed by each agent dominant, generating an authentication and arguments game forest of a multi-agent group, realizing the arguments element organization in the same state space, simultaneously realizing interactive combined antagonism learning through zero and game of a plurality of agents, and generating a difficult disease authentication and diagnosis common arguments set containing positive evidence and reverse evidence. The invention has the autonomy and the interpretability of differential diagnosis evidence-based, not only can provide positive evidence for diagnosing diseases, but also can provide negative evidence for eliminating diseases around the differential doubt points, thereby assisting human doctors in improving the accuracy of differential diagnosis and reducing the requirements on the medical knowledge, experience breadth and depth of the human doctors.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for evidence-based treatment of a problematic disease according to an embodiment of the present invention;
FIG. 2 shows the basic structure and operation logic of the multi-agent differential diagnosis model according to the embodiment of the present invention;
FIG. 3 is a representation of a multi-agent differential diagnosis evidence multi-element set provided by an embodiment of the present invention;
fig. 4 is a joint learning model for constructing a common argument set by the multi-agent system according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, based on the examples herein, which are within the scope of the invention as defined by the claims, will be within the scope of the invention as defined by the claims.
The multi-agent is capable of sensing dynamic environmental information, processing single information through a simple and easy-to-use model, and generating reasoning results to correct decisions. Each agent is specially responsible for a task to make up for the limited ability of a single agent to learn a new task; the system organization and task division make up the limitation of a single agent in the capability of processing information and applying information, and reduce the complexity of the algorithm. UCT is a generic and tree search accuracy that combines stochastic simulation and can be used in the field of defining and predicting output results in { states, actions }, and simulation. Therefore, the invention provides that the multi-agent and UCT algorithm are introduced into medical intelligent differential diagnosis evidence searching, so that the accuracy and efficiency of generating the differential diagnosis basis from the electronic medical record and the knowledge base can be improved.
The invention aims to provide a problematic disease evidence-based method, a system and a storage medium based on an electronic medical record, which can search, organize and represent typical symptoms and similar disease symptoms related to a patient disease from the electronic medical record and provide interpretable positive support evidence and negative exclusion evidence for intelligently and accurately judging the type and the condition of the patient disease. Specifically, the method comprises the steps of firstly constructing a multi-agent differential diagnosis model, collecting clinical information and related examination data of a patient, removing redundant information in the data, and taking the redundant information as input of the multi-agent model. Then, different agents are called for different symptoms, and according to clinical information and examination data of patients, each agent performs simulation prediction of optimal evidence combination through a Monte Carlo method according to an existing knowledge base. And then according to the multi-agent theory, representing elements and representing structural organization, distinguishing and diagnosing the problematic diseases, supporting evidence on the front side and resisting evidence on the back side. The method has the autonomy and the interpretability of differential diagnosis evidence-based, not only can provide positive evidence for diagnosing diseases, but also can provide negative evidence for eliminating diseases around the differential doubt points, thereby assisting a human doctor in improving the accuracy of differential diagnosis and reducing the requirements on the medical knowledge, experience breadth and depth of the human doctor. The technical scheme of the invention is specifically described below with reference to specific embodiments.
As shown in fig. 1, the embodiment of the invention provides a method for evidence-based treatment of a problematic disease based on an electronic medical record, which comprises the following steps:
s1: establishing a multi-agent differential diagnosis model framework: constructing an agent model with a sensing unit, an reasoning unit and an interaction unit, and taking the agent model as a calculation carrier for electronic medical record data processing, differential diagnosis evidence element searching, discussion data organization and representation; the agent model comprises a diagnosis agent and an identification agent; reconstructing a differential-diagnostic challenge process represented by a plurality of agent gaming models. The method specifically comprises the following steps:
s11: an intelligent agent model is constructed based on a JADE (Java Agent Development Framework) platform to serve as a calculation carrier. In general, the multi-agent differential diagnosis model is composed of two functional agents, namely a diagnosis agent and a differential agent, and fig. 2 shows the basic structure and operation logic of the multi-agent differential diagnosis model. The diagnosis agent is mainly responsible for controlling the initialization of the differential diagnosis platform, the initiation of the identification subjects of the difficult and complicated diseases, the management, the control and the result judgment of the identification evidence-based process of the difficult and complicated diseases. The identification agent is an reasoning search evidence-based executor of the identification evidence-based process, executes proper evidence-based construction behavior according to the evidence-based link where the identification agent is located, and has the capability of situation awareness, evidence-based construction and evidence-based interaction.
In this embodiment, to implement the above process, the basic configuration environment of the agent is as follows: JADE platform, OWL (Ontology Web Language), OWAPI+Pellet inference engine, tuProlog core inference engine, etc. JADE is used to develop multi-agent differential diagnosis models and differential diagnosis agent applications conforming to FIPA (The Foundation for Intelligent Physical Agents) standards. The OWL ontology editing tool provides semantic support for agents to understand the meaning expressed by the arguments and mine the attack relationships. The combination of the OWL API and the Pellet inference engine provides standard and pruning inference service for agents to utilize domain ontology to mine attack relations. the tuProlog consists of an interpretation library, an inference engine and a knowledge base, wherein the interpretation library, the inference engine and the knowledge base are internal modules of the 2P engine, and the inference on the basis of the knowledge base is realized; the knowledge base is a set of rules and facts that complete logical reasoning, established by the user.
The intelligent agent sensing unit acquires external environment information required by the intelligent agent, the environment data extracts characteristic information related to disease identification points, and the characteristic information can be used as decision basis for the intelligent agent to perform state adjustment or action execution. For diagnosis/authentication, the agent inference unit learns an inference model conforming to the gist of authentication, and the output result of the model can assist in the generation of the final diagnosis/authentication result. The interaction unit of the intelligent agent is responsible for completing the authentication and discussion data exchange process between the current intelligent agent and other intelligent agents in the system, and the process is an important basis for a plurality of intelligent agents to realize the cooperation of diagnosis and authentication diagnosis and the game of countermeasure and discussion data. Finally, the agent feeds back the final diagnosis/authentication result execution to the environmental information. The above main behavioral logic of diagnosing and identifying multi-agent can be summarized as "sense→inference→diagnosis/identification".
Step 1-2: and establishing a multi-agent differential diagnosis talking gaming framework. In order to realize the extraction, reasoning and calculation of the relevant identification basis (main complaints, current medical history, physical signs, inspection and the like) of the identification key points (the identification key points are different according to different diseases) of the multi-agent (MAS) on the difficult and complicated diseases, a MAS theory game framework is established. The dialect interaction around the authentication gist between different agents is realized by the dialect game, which is expressed as a dialect dialogue game of a diagnosis agent Dia and a plurality of authentication agents Dis-Dia: firstly, dia presents diagnosis focus arguments (namely initial arguments), then Dia and Dis-Dia continuously present effective attack arguments around the focus arguments to attack arguments which have been presented by the opposite party, an allowable set containing the focus arguments is constructed through the interaction of the arguments so as to prove that the diagnosis focus arguments are acceptable or can be abolished, and further, positive evidence for diagnosing the disease and negative evidence for excluding the disease are finally formed through multiple rounds of iteration.
The above process may be formally represented as a multi-agent differential diagnosis talking gaming framework (MAS Argument Game Framework, MAS-AGF):
MAS-AGF＝<DAF, Dia, Dis-Dia, Dis-DiaMDR> (1)
wherein, the DAF stands on the distributed framework of the agent set constructed by the step S11, dia is a diagnostic agent, dis-Dia is an authentication agent set, and Dis-Dia mdr is an interactive game protocol of Dia and Dis-Dia.
The Dis-DiaMDR in the MAS-AGF comprises a control protocol and a game interaction protocol, and the specific steps are as follows:
d1: the diagnostic agent sends an instruction for starting searching evidence to the authentication agent set, and sets the issue as disease;
d2: BROADCAST (Dia, dis-Dia, arg), the diagnostic agent BROADCASTs the primary focus of dispute Arg of the current differential diagnosis to the set of differential agents;
d3: ACCREDIT (Dia, dis-Dia), the diagnostic agent grants authentication agent search rights, meaning the establishment of a new round of evidence-based search;
d4: CONFIRM (Dia, dis-Dia), the diagnostic agent CONFIRMs to all authentication agents whether valid authentication arguments exist, if the authentication agent cannot provide new arguments for changing the state of the current authentication arguments set, the dialectical reasoning process regarding the main dispute focus of the current authentication diagnosis goes into the termination procedure;
d5: DECLARE (Dia, dis-Dia, arg), the diagnostic agent DECLAREs acceptability of the main dispute focus Arg for the current differential diagnosis and terminates the inferential evidence-based process;
d6: CLAIM (i, dis-Dia, arg, evi), discrimination agent i centers around discrimination diagnosis dispute Arg, proposing new evidence Evi;
D7：ARGUE(i,j,Dis-Dia,) Discrimination potential entity i presents a structured organization and representation of the attack theory set +. >To attempt to change the current state of the discussion dataset;
d8: PASS (j, dia, dis-Dia) discrimination agent j returns to diagnostic agent Dia a discrimination diagnosis evidence set giving up its last-stage construction, andreceiving construction of discrimination agent i in D7
Among the above protocols, D1-D5 are control protocols and D6-D8 are game interactive protocols.
Thus, step S11 is completed to construct an agent model having a sensing unit, an inference unit, and an interaction unit as a calculation carrier for electronic medical record data processing, differential diagnosis evidence element searching, discussion organization, and representation. Step S12 establishes a differential diagnosis multi-agent discussion gaming framework of a differential-diagnosis countermeasure process model represented in an interactive game of a plurality of agents.
S2: patient electronic medical record association history electronic medical record screening. Clinical information and relevant examination data of a patient are collected, and similar associated historical medical records are searched and screened in an electronic medical record database and a knowledge base according to association rules of patient complaints, medical history, abnormal signs and examination and inspection items.
S21: the method comprises the steps of extracting characteristics of an electronic medical record, defining terms such as human body parts, disease names, symptoms, inspection and examination items, operations, treatments and the like described by natural language in the electronic medical record as medical named entities, performing self-attention calculation and residual calculation of a plurality of layers of encoders of a transducer, taking output of a topmost Encoder unit as input of a conditional random field (Conditional Random Fields, CRF) information extraction algorithm, obtaining label classification of the named entities, and performing entity identification and extraction. The method is mainly used for screening and extracting historical electronic medical record data with correlation according to issue diseases initiated by Dis-DiaMDR in MAS-AGF, and the historical electronic medical record data is used as input data and a knowledge base for identifying the Dis-Dia of the agent.
The specific characteristic extraction process is as follows: the data set identified by the medical concept entity is marked, and in the marking of the data set, some common medical concepts and relations between the medical concepts are mainly marked. A transducer-based bi-directional encoder model is trained from a corpus of the disclosed medical domain such that the model can identify entities and relationships in medical text. Then using the modelPredicting a medical text data set without labels, and identifying medical concepts in the medical text and relations between the medical concepts. Obtaining topics of the electronic medical record by using a document topic generation model (three-layer Bayesian probability model) with a word, topic and document three-layer structure, and obtaining the distribution probability P (z) of each topic word of the electronic medical record through the document topic generation model r )：
Wherein z is r The representative topics r, S are the number of topics (s= … S), and V is the number of words (v= … V); alpha is the hyperparameter of the directlet of the theme, the subscript t represents the word t, and v is all words; beta is the directlet super parameter of the word, the subscript k represents the topic k, and s is all topics;is the number of times a word t is assigned to a topic k other than the current topic,/>Is the total number of words assigned to topic k; />For the number of times that topic k is assigned to document m other than the current document, +. >Is the total number of topics assigned to document m.
Finally, N subject words with highest probability (such as 4, 5, 6 and the like) are selected as the subjects of the electronic medical record document.
S22: the historical electronic medical record database and the knowledge base are associated with the historical electronic medical record retrieval. Firstly, scanning a historical electronic medical record database and a knowledge base once based on a document theme of the electronic medical record of a patient, and counting the occurrence times of corresponding subject words in each historical electronic medical record to form a first candidate item set; screening a first frequent item set with association rules meeting minimum strength according to a minimum support threshold; combining the first frequent item sets, namely combining the subject words in the first frequent item sets in pairs to form a second candidate item set; performing second scanning on the historical electronic medical record database and the knowledge base, counting each second candidate item set according to the first candidate item set counting method, and screening out second frequent item sets according to the minimum support threshold; performing iterative screening according to the scanning and screening method, namely sequentially increasing the number of the combined subject words to perform scanning and screening until the candidate item set is empty; and according to the finally generated frequent item set, generating the association between the electronic medical record of the patient and the historical electronic medical record by calculating the similarity between the historical electronic medical record and the electronic medical record of the patient in the finally generated frequent item set, removing the electronic medical record with the similarity lower than a similarity threshold value, and obtaining the candidate set of the identification agent search and the generation of the identification diagnosis evidence.
The medical record similarity calculation uses a graph attention (drawing attention mechanism) improved transducer model to calculate the medical record similarity according to the electronic case information input by the user. The method comprises the following specific steps:
inputting the text and the theme of the electronic medical records in the training data into a transducer model, and calculating the similarity between the electronic medical records by using a twin neural network sharing weights. Firstly, each word in the electronic medical record input by the user is converted into a word embedding vector through an embedding layer by using the obtained word embedding vector. The output of the embedded layer is then input into a transducer model that shares weights. The obtained word embedding vector is added into a transducer network to enhance the effect of a model, the training effect is enhanced by using a attention mechanism through data supervision and learning with labels, the model is trained, and finally a medical record similarity calculation model is obtained through parameter adjustment and model optimization and is used for screening the patient medical record association history electronic medical record.
In order to realize multi-agent autonomous screening, a trained model is embedded into a model library of diagnostic agents Dia, and when a user inputs patient case information, the Dia automatically invokes the model screening and generates an identification agent set Dis-Dia to search for an alternative set of identification diagnosis evidence elements.
S3: the agent searches for evidence. And searching and expanding a UCT tree by adopting a UCT searching algorithm to realize the automatic generation of an optimal identification and discussion set. The authentication agent acquires and evaluates an optimal authentication argument game tree (authentication argument UCT tree) of the current authentication gist by repeating the UCT search a plurality of times. Firstly, a path reaching a leaf node (identification theory formed by a historical electronic medical record) is selected by a random strategy from a root node (identification theory of a problematic disease) of an identification theory game tree, and the obtained leaf node is expanded by utilizing MCTS (Monte Carlo Tree Search) Monte Carlo tree search simulation MAS benefits as a game tree node expansion basis, then Monte Carlo simulation game results are carried out on the leaf node and recorded, and finally, the simulated results are updated to the node values according to the path. The method comprises the following steps:
s31: authentication arguments UCT tree initialization. Since the search aims at taking patient electronic case data input by a user as a root, searching evidence which is valuable for the current diagnosis in the historical electronic medical record as a diagnosis basis. Thus, the data structure of the tree structure is selected for searching. The identification agent judges the possible diseases in the patient electronic case input by the user and judges the main points of the identification needed by the diseases, the identification of the initial state is based on the root node of the UCT tree, and the root node stores the electronic medical record data of the current patient, comprising: patient complaints, current medical history, abnormal signs, inspection and inspection index values, etc., are used as the status of the current stage of the discrimination agent.
S32: discrimination accounts for the selection and expansion of the UCT tree. The authentication agent selects the child node according to the upper bound confidence algorithm UCB (upper confidence bound), and adopts a strategy of balancing the known value and the unknown value to select the child node. The child node stores a historical electronic medical record which is screened according to the step 2 and is related to the current patient case input by the user, wherein the historical electronic medical record comprises the following components: patient complaints, current medical history, abnormal signs, inspection and inspection index values, and the like. The UCB algorithm calculates the confidence upper bound for each child node, which contains two factors: known value and unknown value. The known value represents the average return of the child nodes and the unknown value represents the exploratory value of the child nodes. The confidence assessment calculation formula is as follows:
wherein,representing the average benefit of agent i in the current search state; n is n i Indicating the number of searches for agent i, and n indicating the number of searches for all agents.
And evaluating each historical electronic medical record sub-node according to the formula, and selecting the sub-node with the maximum confidence upper bound to update the current authentication and discussion UCT tree.
S33: discrimination accounts for the simulation and backtracking of the UCT tree. The method comprises the steps of developing from extended sub-nodes, performing simulation evaluation by using an OWL API+Pellet inference engine of an identification agent, calculating the characteristics of patient complaints, current medical history, abnormal signs, inspection and inspection index values and the like in a historical electronic medical record, and the relation between the characteristics of the current patient case, updating the access times, accumulated return values and average return values of each node of the whole UCT tree by the sub-nodes to an upper root node, pruning the UCT tree according to return evaluation values to form a new identification data UCT tree, wherein the identification data UCT tree is formed by the root node of the current patient case, the post-pruning historical electronic medical record sub-nodes and the support relation return evaluation values among the nodes.
S34: the above steps S32 and S33 are repeated until the authentication argument UCT tree estimation converges or a preset termination condition such as the upper limit of the search time or the number of times is reached. And obtaining a final authentication argument UCT tree, namely storing the optimal authentication argument set of the historical electronic medical record.
S4: and carrying out structural organization and representation of the differential diagnosis evidence according to the differential arguments UCT tree. Because the UCT is a random search strategy, the structure of UCT trees constructed by different agents has variability, and the identification interaction process between the agents is difficult to realize. Therefore, the Toulmin theory is used for organizing theory elements, and the theory expression with uniformity, integrity and structure is constructed for the theory interaction of multiple agents.
In order to realize MAS diagnosis and differential diagnosis instantiation based on the discussion game, the discussion elements and the discussion structure based on the electronic medical record data are required to be constructed by combining the disease identification key points. Toulmin proposes a six-tuple model of the arguments structure, called Toulmin model, and performs structural processing on the arguments, which is beneficial to MAS management and interaction of the arguments. The arguments represented by the classical Toulmin model consist of claims (Claim), data (Data), arguments (Warrant), support (backswing), modifier (modifier) and refute (refute), in order to accommodate the construction of the arguments around the gist of difficult disease discrimination by MAS, the present embodiment improves the foregoing arguments structure model to:
Arg-Evi＝{Diagnosis，Differential，Warrant，Backing，Strength，Data}(4)
Wherein, diagnostis represents the assertion of the focus of Diagnosis of problematic diseases and similar diseases of diagnostic agents; difference represents an authentication evidence (reverse evidence) element of an authentication agent, which is a feature extracted from an electronic medical record of a patient stored in an authentication and argumentation UCT tree; warrant represents the mapping relation of electronic medical record data to authentication evidence elements of an authentication agent; backsight represents identifying supplementary evidence elements stored by sub-nodes in the UCT tree as secondary arguments; strength represents the Strength estimate of the Differential, i.e., the return estimate in step S33; data represents the original Data in the historical electronic medical record and the current electronic medical record of the patient, and is obtained from the alternative set selected in the step 2-2.
The specific Arg-Evi theory structure is shown in figure 3.
S5: the group agents perform joint interactive learning based on UCT trees and Arg-Evi argument structure models of the individual agents, and a common argument set is constructed. The hidden tree constructed by the UCT process of each agent is made explicit, and a discrimination and discussion forest (common discussion set) of the MAS group is generated, so that the MAS can conveniently conduct knowledge and data interaction, and MAS diagnosis and discrimination diagnosis of difficult and complicated diseases in the same state space are supported. The specific implementation process comprises the following steps:
After the Agent builds a UCT-based authentication and discussion game tree (step S3), the implicit discussion of the Agent of the individual is made dominant, and EAS (Experience Argument Schema) is used as an element for authentication and diagnosis; the Arena model in the discussion game is adopted to carry out joint learning (see figure 4), interaction and comparison of the discussion among agents are realized, the winning discussion after multiple rounds of games is used as the optimal global discussion for the current identification key point which is superior to other agents, a common discussion set is achieved through calculation, and local discussion is extracted and stored as group discussion; and finally, taking the MAS common arguments as a common diagnosis and differential diagnosis decision basis set, performing spiral evolution of the arguments game tree, and generating a multi-agent arguments game forest as a differential diagnosis basis.
Wherein the Arena joint learning model is as follows:
wherein, disease is the issue of differential diagnosis, dia is the diagnostic agent, dis-Dia is the set of differential agents, arg is the focus of differential diagnosis disputes, UCT-Tree is the set of differential data UCT Tree constructed by the agents,for the attack theory data set of structural organization and representation, MAS-AGF is a differential diagnosis multi-agent theory game framework, GK is a high-quality global knowledge set which is obtained by the multi-agent through joint interactive learning aiming at the differential diagnosis dispute focus Arg in the current subject treatment.
The MAS group common arguments set up process combined with Arena joint learning model is as follows: the diagnostic agent Dia broadcasts to the authentication agent set Dis-Dia according to the break and Arg input by the user, and Dis-Dia realizes role conversion and multiparty discussion game according to MAS-AGF so as to eliminate inconsistent evidence in UCT-Tree, specifically throughConflict resolution and consistency fusion are carried out on evaluation inconsistent evidence of (a) to obtain a consensus-reached claim and the consensus-reached claimDiagnosis conclusion supported. The final winning Dis-Dia (i) comprises +.>The method is a high-quality global knowledge obtained by multi-agent joint learning.
All the identification intelligent agents repeatedly complete the joint learning process to obtain a global knowledge set (common discussion data set) GK related to the current patient case disease, namely, the identification diagnosis evidence of the possible diseases, which is searched and generated in the historical electronic medical record library by MAS aiming at the current patient case disease, and the identification intelligent agents can assist human doctors to realize the identification diagnosis process of the diseases. The method can make up the knowledge blind area of the individual doctors, realize the sharing of the knowledge of the electronic medical record generated by the diagnosis of all doctors, and improve the accurate acquisition of the differential diagnosis of the complex diseases. In particular to a method for searching and discovering global knowledge and diagnosis evidence aiming at interdisciplinary diagnosis of complex diseases.
The evidence-based method for treating the difficult and complicated diseases provided by the embodiment has the following beneficial effects:
1. the human doctor differential diagnosis logic is introduced into the AI auxiliary diagnosis, a set of intelligent differential diagnosis modes of the difficult and complicated diseases containing the evidence-based process are established, the interpretation is strong, and the doctor is assisted to promote the accuracy of the difficult and complicated disease diagnosis.
2. Inspired by a machine game algorithm, the multi-agent and UCT algorithm is used for simulating the evidence-based process of diagnosis-identification of a human doctor, and compared with a deep learning algorithm, a visual evidence tree can be generated to explain decisions, so that more focused and accurate positive evidence and reverse evidence can be obtained.
3. The evidence-based process is realized by using the multi-agent system with sensing, reasoning and interaction functions, the distributed computation generates evidence, and compared with a complex large model, the evidence-based process is simpler and more easy to use by using the agent model, and the intelligent-based effect of the multi-agent distributed computation, interaction and collaborative reasoning ensures the algorithm effect while improving the algorithm efficiency.
The embodiment of the invention also provides a system for evidence-based treatment of the difficult and complicated diseases, which comprises:
a memory having a computer program stored thereon;
and the processor is used for realizing the evidence-based method for the problematic diseases based on the electronic medical record according to the embodiment when the computer program is loaded and executed.
The embodiment of the invention also provides a computer readable storage medium, on which a computer program is stored, which when being executed by a processor, realizes the method for evidence-based on the electronic medical record.
It is to be understood that the same or similar parts in the above embodiments may be referred to each other, and that in some embodiments, the same or similar parts in other embodiments may be referred to.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.
Claims (10)
1. An electronic medical record-based evidence-based method for evidence-based treatment of difficult and complicated diseases is characterized by comprising the following steps:
s1: establishing a multi-agent differential diagnosis model framework, constructing a plurality of agent models including diagnosis agents and differential agents, and representing a differential-diagnostic countermeasure process model by using zero and game models of the plurality of agents;
s2: acquiring an electronic medical record of a patient, and searching and screening similar associated historical electronic medical records in a historical electronic medical record database and a knowledge base by a diagnosis agent;
s3: the authentication agent adopts UCT searching algorithm to obtain and evaluate authentication argument UCT tree of the current authentication key point in the screened similar association history electronic medical record;
s4: according to the authentication argument UCT tree, carrying out structural organization and representation of the authentication argument for the argument interaction of multiple intelligent agents;
s5: the implicit identification theory data UCT tree constructed by each agent is made explicit to generate an identification theory game forest of a multi-agent group, thereby realizing theory element organization in the same state space, and simultaneously realizing interactive combined antagonism learning through zero and game of a plurality of agents to generate a difficult disease identification diagnosis common theory data set containing positive evidence and negative evidence.
2. The method according to claim 1, wherein the step S1 comprises:
the diagnosis agent and the identification agent are agent models comprising a sensing unit, an reasoning unit and an interaction unit; the diagnosis agent is responsible for controlling the initialization of differential diagnosis, the initiation of the identification issue of the problematic disease, the management, the control and the result judgment of the identification evidence-based process of the problematic disease; the identification agent is an reasoning search evidence-based executor for identifying the evidence-based process, and executes the discussion data construction behavior according to the evidence-based link where the evidence-based executor is located;
the discrimination-diagnostic countermeasure process model is represented in a zero and game model of a plurality of agents, which is represented as a talking dialogue game of a diagnostic agent and a plurality of discriminating agents: firstly, the diagnosis focus theory is provided by the diagnosis intelligent agent, then the diagnosis intelligent agent and a plurality of identification intelligent agents continuously provide attack theory around the diagnosis focus theory to attack theory provided by the opposite party, and an allowable set containing the diagnosis focus theory is constructed through theory interaction so as to prove that the diagnosis focus theory is acceptable or can be abolished.
3. The method according to claim 1, wherein the step S2 comprises:
S21: extracting the characteristics of the electronic medical record of the patient, and the electronic medical records of the history electronic medical record database and the knowledge base to obtain the distribution probability of each subject term of the electronic medical record, and selecting N subject terms with the highest probability as the document subjects of the corresponding electronic medical record, wherein N is a preset value;
s22: the method comprises the steps of searching associated historical medical records, firstly, scanning a historical electronic medical record database and a knowledge base once based on a document theme of the patient electronic medical record, and counting the occurrence times of corresponding subject words in each historical electronic medical record to form a first candidate set; screening a first frequent item set with association rules meeting minimum strength according to a minimum support threshold; combining the first frequent item sets to form a second candidate item set; performing second scanning on the historical electronic medical record database and the knowledge base, counting each second candidate item set according to the first candidate item set counting method, and screening out second frequent item sets according to the minimum support threshold; iterative screening is carried out according to the scanning and screening method until the candidate item set is empty; and according to the finally generated frequent item set, generating the association between the electronic medical record of the patient and the historical electronic medical record by calculating the similarity between the historical electronic medical record and the electronic medical record of the patient in the finally generated frequent item set, removing the electronic medical record with the similarity lower than a similarity threshold value, and obtaining the candidate set of the identification agent search and the generation of the identification diagnosis evidence.
4. The method according to claim 1, wherein the step S21 comprises:
identifying and extracting entities and relations in the electronic medical record by using a trained bidirectional encoder model based on a transducer;
and obtaining the distribution probability of each entity or relation subject word of the electronic medical record by using the document subject generation model, and selecting N subject words with the highest probability as the document subjects of the corresponding electronic medical record.
5. The method according to claim 1, wherein the step S3 comprises:
s31: authentication arguments UCT tree initialization: taking the identification key points of the problematic diseases as the root nodes of the UCT tree in the initial state, wherein the root nodes store the electronic medical record data of the current patient;
s32: discrimination argues about the selection and expansion of the UCT tree: calculating the confidence upper bound of each child node according to the UCB algorithm, and selecting the child node with the maximum confidence upper bound for expansion so as to update the current authentication argument UCT tree; each child node stores a similar association history electronic medical record obtained by screening;
s33: discrimination arguments are the simulation and backtracking of the UCT tree: the method comprises the steps of developing from expanded sub-nodes, performing simulation evaluation by using an inference engine, and specifically calculating the relation between the characteristics of a historical electronic medical record and the characteristics of patient complaints, current medical history, abnormal physical signs and check indexes in the current electronic medical record of a patient, updating the access times, accumulated return values and average return values of each node of the whole authentication and discussion UCT tree from the sub-nodes to a root node, pruning the authentication and discussion UCT tree according to the return evaluation values to form a new authentication and discussion UCT tree, wherein the authentication and discussion UCT tree consists of the root node of the current electronic medical record of the patient, the sub-nodes of the history electronic medical record after pruning and the return evaluation values of the supporting relation among the nodes;
S34: and repeating the steps S32 and S33 until the evaluation value of the authentication and discussion data UCT tree converges or the preset termination condition of the upper limit of the searching time or the upper limit of the times is reached, and obtaining the final authentication and discussion data UCT tree, namely storing the optimal authentication and discussion data set of the historical electronic medical record.
6. The method according to claim 1, wherein in step S4, the identification argument elements are organized in six tuples with Toulmin argument to construct a complete identification argument representation; the discriminant theory structure Arg-Evi is represented as follows:
Arg-Evi＝{Diagnosis，Differential，Warrant，Backing，Strength，Data}
wherein diagnostis represents the proposition of a focus of Diagnosis of problematic diseases and similar diseases of diagnostic agents; the Differential represents an identification agent identification evidence element, which is the characteristic extracted from the patient electronic medical record stored in the identification data UCT tree; warrant represents the mapping relation of electronic medical record data to authentication evidence elements of an authentication agent; backsight represents identifying supplementary evidence elements stored by sub-nodes in the UCT tree as secondary arguments; strength represents the intensity estimate of the Differential; data represents raw Data in the historic electronic medical record and the current patient electronic medical record.
7. The method according to claim 1, wherein the step S5 comprises:
After constructing the authentication argument UCT tree, the implicit authentication argument of each agent is made dominant, and the implicit authentication argument is expressed as an element for differential diagnosis in an EAS empirical proof mode;
adopting an Arena joint learning model in zero and game to perform joint learning, realizing interaction and comparison of arguments among agents, taking the winning arguments after multiple rounds of game as optimal global arguments for the current identification key points, which are superior to other agents, achieving a joint arguments set through calculation, and extracting local arguments to be stored as group arguments;
the multi-agent common arguments are used as a common diagnosis and differential diagnosis decision basis set to carry out spiral evolution of the arguments game tree, and a multi-agent arguments game forest is generated to be used as the differential diagnosis basis.
8. The method for evidence-based on electronic medical records of claim 7, wherein the Arena joint learning model is as follows:
wherein, disease is the issue of differential diagnosis, dia is the diagnostic agent, dis-Dia is the set of differential agents, arg is the focus of differential diagnosis disputes, UCT-Tree is the set of differential data UCT Tree constructed by the agents,for the attack theory data set of structural organization and representation, MAS-AGF is a differential diagnosis multi-agent theory game framework, GK is a high-quality global knowledge set which is jointly and interactively learned by the multi-agent aiming at the differential diagnosis dispute focus Arg in the current subject treatment;
Multi-intelligence combined with Arena joint learning modelThe process of building the volume common arguments is as follows: the diagnosis agent broadcasts to the identification agent set according to the events and Arg input by the user, and the identification agent set realizes role conversion and multiparty discussion game according to MAS-AGF so as to eliminate inconsistent evidence in UCT-Tree, in particular byCarrying out conflict resolution and consistency fusion on the estimated value inconsistent evidence to obtain a consensus-achieved claim and a diagnosis conclusion supported by the consensus-achieved claim; the final winning discrimination agent comprises ++>The method is a high-quality global knowledge obtained by multi-agent joint learning;
and repeatedly completing the joint learning process by all the identification intelligent agents to obtain a high-quality global knowledge set GK related to the current patient case event as a common argumentation set, namely, the identification diagnosis evidence of the possible diseases, which is searched and generated in a historical electronic medical record library and a knowledge base by the multi-intelligent agent aiming at the current patient case event.
9. A problematic disease evidence-based system, comprising:
a memory having a computer program stored thereon;
a processor for implementing the electronic medical record-based evidence-based method of problematic disease as claimed in any one of claims 1 to 8 when said computer program is loaded and executed.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the method for evidence-based on electronic medical records of the problematic disease according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311276855.2A CN117352159A (en) | 2023-09-28 | 2023-09-28 | Method, system and storage medium for evidence-based treatment of difficult and complicated diseases based on electronic medical records |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311276855.2A CN117352159A (en) | 2023-09-28 | 2023-09-28 | Method, system and storage medium for evidence-based treatment of difficult and complicated diseases based on electronic medical records |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117352159A true CN117352159A (en) | 2024-01-05 |
Family
ID=89366024
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311276855.2A Pending CN117352159A (en) | 2023-09-28 | 2023-09-28 | Method, system and storage medium for evidence-based treatment of difficult and complicated diseases based on electronic medical records |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117352159A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117577348A (en) * | 2024-01-15 | 2024-02-20 | 中国医学科学院医学信息研究所 | Identification method and related device for evidence-based medical evidence |
| CN117995392A (en) * | 2024-04-07 | 2024-05-07 | 北京惠每云科技有限公司 | Differential diagnosis generation method, device, electronic equipment and storage medium |
-
2023
- 2023-09-28 CN CN202311276855.2A patent/CN117352159A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117577348A (en) * | 2024-01-15 | 2024-02-20 | 中国医学科学院医学信息研究所 | Identification method and related device for evidence-based medical evidence |
| CN117577348B (en) * | 2024-01-15 | 2024-03-29 | 中国医学科学院医学信息研究所 | Identification method and related device for evidence-based medical evidence |
| CN117995392A (en) * | 2024-04-07 | 2024-05-07 | 北京惠每云科技有限公司 | Differential diagnosis generation method, device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tripathy et al. | A framework for intelligent medical diagnosis using rough set with formal concept analysis | |
| US20210233658A1 (en) | Identifying Relevant Medical Data for Facilitating Accurate Medical Diagnosis | |
| Tillers et al. | A theory of preliminary fact investigation | |
| CN117352159A (en) | Method, system and storage medium for evidence-based treatment of difficult and complicated diseases based on electronic medical records | |
| CN106295186A (en) | A kind of method and system of aided disease diagnosis based on intelligent inference | |
| Gharehchopogh et al. | Neural network application in diagnosis of patient: a case study | |
| CN110838368A (en) | Robot active inquiry method based on traditional Chinese medicine clinical knowledge graph | |
| CN109949929A (en) | A kind of assistant diagnosis system based on the extensive case history of deep learning | |
| Grando et al. | Argumentation-logic for creating and explaining medical hypotheses | |
| Yang et al. | Predicting coronary heart disease using an improved LightGBM model: Performance analysis and comparison | |
| Bhatia et al. | Prediction of severity of diabetes mellitus using fuzzy cognitive maps | |
| CN116741411A (en) | Intelligent health science popularization recommendation method and system based on medical big data analysis | |
| Chou et al. | Extracting drug utilization knowledge using self-organizing map and rough set theory | |
| Prakken et al. | Editors' review and introduction: Models of rational proof in criminal law | |
| CN110610766A (en) | Apparatus and storage medium for deriving probability of disease based on symptom feature weight | |
| CN114496231B (en) | Knowledge graph-based constitution identification method, device, equipment and storage medium | |
| Sameh | Knowledge based bayesian network construction algorithm for medical data fusion to enhance services and diagnosis | |
| Savita et al. | An Optimized Machine Learning Approach for Coronary Artery Disease Detection | |
| Kong | An online belief rule-based group clinical decision support system | |
| Tesfaye | Developing a Learning Knowledge-Based System For Diagnosis And Treatment Of Malaria | |
| CN110289065A (en) | A kind of auxiliary generates the control method and device of medical electronic report | |
| Qiu et al. | Building Syndrome and Symptom Association Network by Bayesian Network | |
| Kahsu | A CASE BASED REASONING SYSTEM FOR DIAGNOSIS OF MALNUTRITION FOR UNDER-FIVE YEAR CHILDREN: THE CASE OF TIRUNESHE BEJING | |
| Tian et al. | Weighted Heterogeneous Graph-Based Incremental Automatic Disease Diagnosis Method | |
| Salem et al. | Exploiting the knowledge computing and engineering in medical informatics and healthcare |
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 |