CN114496234B - Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients - Google Patents

Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients Download PDF

Info

Publication number
CN114496234B
CN114496234B CN202210403938.2A CN202210403938A CN114496234B CN 114496234 B CN114496234 B CN 114496234B CN 202210403938 A CN202210403938 A CN 202210403938A CN 114496234 B CN114496234 B CN 114496234B
Authority
CN
China
Prior art keywords
patient
node
cognitive
data
map
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210403938.2A
Other languages
Chinese (zh)
Other versions
CN114496234A (en
Inventor
李劲松
刘强华
田雨
周天舒
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN202210403938.2A priority Critical patent/CN114496234B/en
Publication of CN114496234A publication Critical patent/CN114496234A/en
Application granted granted Critical
Publication of CN114496234B publication Critical patent/CN114496234B/en
Priority to PCT/CN2023/088590 priority patent/WO2023202508A1/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G06N5/022—Knowledge engineering; Knowledge acquisition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • 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)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Primary Health Care (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Pathology (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The invention discloses a general patient personalized diagnosis and treatment scheme recommendation system based on a cognitive atlas, which comprises a data acquisition module, a data preprocessing module, a data analysis reasoning module and a recommendation result display module, wherein a general knowledge atlas is firstly constructed in the data analysis reasoning module, and then the cognitive atlas of a patient personalized disease development track is established based on information such as diseases, symptoms and medicines when the patient visits and the constructed general knowledge atlas, so that the personalized diagnosis and treatment scheme recommendation of the patient is given. The invention uses the inference method based on the cognitive map, so that the system can truly simulate the diagnosis and treatment thought of a clinician, and provides an interpretable clinical auxiliary decision tool with high acceptability for the clinician; the invention sets up a personalized diagnosis and treatment scheme for the patient from symptoms, helps the patient to discover the cause of disease as soon as possible and receive targeted treatment, and can realize early screening of dangerous diseases and prompt the patient to make a referral to specialized treatment in time.

Description

Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients
Technical Field
The invention belongs to the technical field of medical health information, and particularly relates to a cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients.
Background
In China, general medicine is a comprehensive medical professional subject which is oriented to communities and families and integrates relevant contents of clinical medicine, preventive medicine, rehabilitation medicine and human social subject, is established in the nineties of the last century, and covers various ages, sexes, various organ systems and various health problems. The main service field of general medicine is primary health care, which mainly deals with common problems in the background of families and communities, and a lot of health problems are in the undifferentiated stage of diseases. A global undifferentiated disease refers to a medically unexplained physical condition or to a disease in which the disease has not been definitively assigned to a certain system in its early stages. That is, patients in the general clinic often show only explicit symptoms, and may not know the cause of their disease and the system to which the disease belongs. In addition, many diseases often show some specific and non-specific symptoms at an early stage of development, such as abdominal pain affecting almost all abdominal cancers, including colon, prostate, bladder, and kidney cancers, and Lancet Oncology in 2020 also indicates that some symptoms in many patients are present during stages I-III of the cancer. To sum up, it is the skills that the general practitioner should learn and master to treat undifferentiated diseases scientifically and to diagnose the diseases at an early stage. However, although the general practitioner culture system has been developed primarily, the culture mode has been established basically, and the number of teams is increasing, the general practitioner number is still insufficient and the quality needs to be improved. In view of the above situation, a patient personalized diagnosis and treatment scheme recommendation system facing global undifferentiated diseases is constructed, which can help global doctors to realize more accurate diagnosis and treatment and early referral, optimize diagnosis and treatment processes, and possibly improve the timeliness of cancer diagnosis and the prognosis of cancer.
At present, personalized diagnosis and treatment scheme recommendation systems or auxiliary diagnosis systems aiming at general patients starting from symptoms are few, the field is still in a development stage, and the prior similar technical schemes have the following defects:
1. in the prior art, few patients who may suffer from different types or different systems of diseases are analyzed by taking symptoms as starting points, most of the diseases are concentrated on one or one type of diseases which are already determined, such as diabetes, autism, infectious diseases and the like, but in the general outpatient service, doctors face the undifferentiated diseases, and the patients who show the same symptoms may need to be transferred to different departments for subsequent treatment, so the current technical scheme cannot well solve the problems encountered in the general field at present;
2. the prior art scheme is mainly used in the process of guide diagnosis or pre-inquiry before the patient and the doctor make a diagnosis and the self-monitoring of the patient at home, is not embedded in the real-time communication activity with the doctor and cannot be really applied to the clinical scene;
3. the existing technical scheme basically adopts a data-driven method, extracts a large number of patient queues meeting conditions, then uses a traditional machine learning or deep learning method to analyze based on clinical data of patients before diagnosis, assists doctors to carry out clinical decision support, does not integrate clinical guidelines, expert consensus and the like in the whole process, and lacks knowledge drive;
4. in the prior art, model training and data analysis are often performed based on big data of a certain crowd, the model effect is greatly reduced after the crowd queue is replaced, namely the generalization performance of the model is poor, meanwhile, the problem of 'black box' exists in deep learning, and the model has no interpretability or poor interpretability.
Disclosure of Invention
The invention aims to provide a cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients, which has high expandability, interpretability and generalization and can solve the following technical problems:
1. according to the technical scheme, aiming at the problem of the undifferentiated diseases of the whole family, a plurality of common clinical symptoms are selected, the symptoms can be radiated to different types of clinical diseases, and personalized support is provided for the diagnosis and treatment scheme of a patient by taking the symptoms as a starting point;
2. the technical scheme can independently provide clinical decision support for the whole diagnosis process of the patient and can also interact with the doctor in real time, the doctor can add new diagnosis or treatment suggestions in the reasoning process at any time, and the system provided by the invention and the doctor together provide more accurate decision support for the patient;
3. the technical scheme adopts a mode of combining data drive and knowledge drive, firstly, a full-family knowledge map is constructed by using a semi-automatic method, then a patient queue meeting conditions is selected according to common clinical symptoms, and finally, a patient personalized diagnosis and treatment scheme recommendation system is constructed by using the full-family knowledge map and a selected data set based on the thought of the knowledge map;
4. the technical scheme simulates the reasoning idea of human solving problems, uses the reasoning method based on the cognitive map, does not limit the number of input variables, has good expandability, realizes the mode of reasoning by combining domain knowledge, has higher generalization performance, can visually display the reasoning process, and has strong interpretability.
The purpose of the invention is realized by the following technical scheme: a general patient personalized diagnosis and treatment scheme recommendation system based on a cognitive map comprises:
a data acquisition module: the system is used for extracting clinical diagnosis and treatment data of the general outpatient clinic from a medical institution database;
a data preprocessing module: carrying out data preprocessing operation on the data acquired by the data acquisition module, including the structured processing and data normalization of the text data of the electronic medical record of the outpatient department;
the data analysis reasoning module: constructing a general knowledge graph, establishing a cognitive graph of a personalized disease development track of a patient by utilizing a variant graph neural network and an attention capture module based on diseases, symptoms and medication information of the patient during the visit and the constructed general knowledge graph, and recommending a personalized diagnosis and treatment scheme of the patient;
a recommendation result display module: the reasoning process of the data analysis reasoning module is visually displayed, and the personalized disease development and treatment process of the patient is provided.
Furthermore, in the data preprocessing module, the symptoms, illness records and medication records of the patient are extracted from the outpatient electronic medical record text data by adopting a method based on rules and a domain dictionary, so that the text data structuring is realized;
only data with positive index results are reserved according to the data types of the examination results containing numerical values; removing patient data from regular medical institution prescription, regular physical examination and regular review; remove the data of the patients in the clinic of traditional Chinese medicine.
Further, in the data analysis reasoning module, the general knowledge graph is constructed in a mode of combining manual construction and automatic construction; the clinical doctor completes the manual construction part of the general knowledge map of the symptom and disease related knowledge extracted by the data acquisition module; the automatic construction part of the general knowledge map covers relevant clinical terms and relations extracted from the SemMed DB and Chinese ICD-10.
Further, the data analysis reasoning module comprises a general knowledge graph construction module and a reasoning system based on a cognitive graph; the cognitive map-based reasoning system comprises two subsystems: the subsystem 1 is used for carrying out disease development track exploration on the constructed full-knowledge map based on the initial clinical diagnosis and treatment data set of the patient; the subsystem 2 is used for accurately positioning the final state of the patient based on a plurality of potential disease development tracks of the patient explored by the subsystem 1.
Further, the subsystem 1 comprises: carrying out global vectorization on the general knowledge graph; the individualized cognitive map of the patient is initially formed by nodes formed by a plurality of initial patient data in an initial clinical diagnosis and treatment data set of the patient, and a potential cognitive map inference node set is iteratively captured on the basis of the cognitive map and a global vectorized general knowledge map;
the subsystem 2 comprises: and gradually supplementing the nodes in the cognitive map inference node set into the cognitive map, defining a node state updating mode of the cognitive map, and inferring and obtaining diseases and treatment modes of the patient based on the general knowledge map and the cognitive map.
Further, in the subsystem 1, global vectorization of the general knowledge graph is realized based on the variogram neural network, specifically:
the global knowledge graph comprises a series of triplets
Figure 902584DEST_PATH_IMAGE001
Each triplet consisting of a relationshiprAnd two entities
Figure 757408DEST_PATH_IMAGE002
Each entity is used as a node in a general knowledge graph, and each triple represents a piece of medical knowledge;
for each node of the global knowledge graphhDefining a learnable embedded vector representation
Figure 3057DEST_PATH_IMAGE003
For each relationshiprDefining a learnable embedded vector representation
Figure 601528DEST_PATH_IMAGE004
(ii) a Setting global vectorization process iteration stepsMFirst, of
Figure 884742DEST_PATH_IMAGE005
While iterating, the nodehIs represented as a state vector of
Figure 972784DEST_PATH_IMAGE006
，
Figure 708659DEST_PATH_IMAGE007
And nodehRelated medical knowledge
Figure 110821DEST_PATH_IMAGE001
Vectorized representation of
Figure 248542DEST_PATH_IMAGE008
；
Pairing nodes in a holistic knowledge graphhAggregating medical knowledge contained in all the triples to obtain a triplet aggregation vector
Figure 710747DEST_PATH_IMAGE009
Wherein
Figure 730655DEST_PATH_IMAGE010
Is a node in the general knowledge graphhThe number of all triplets in which it resides;
node pointhOf the next step state vector
Figure 936509DEST_PATH_IMAGE011
In an update manner of
Figure 663156DEST_PATH_IMAGE012
；
Figure 561842DEST_PATH_IMAGE013
And
Figure 69047DEST_PATH_IMAGE014
the method comprises the steps that two hidden layer functions are adopted, the input is the splicing of all vectors, and the output is consistent with the dimension of the input single vector;
the output result of the last step
Figure 813012DEST_PATH_IMAGE015
As nodeshIs represented by a global vector.
Further, in the subsystem 1, a potential cognitive atlas inference node set is iteratively captured from a globally vectorized general knowledge atlas based on an initial clinical diagnosis and treatment data set through an attention capture module, specifically:
defining node access sets
Figure 659745DEST_PATH_IMAGE016
Initial node Access Collection
Figure 463753DEST_PATH_IMAGE017
S is the initial clinical data set of the patient, will
Figure 458254DEST_PATH_IMAGE018
Each node in the set is marked as not-accessed, and the maximum iteration step number is setLWhen the number of iteration steps is
Figure 5910DEST_PATH_IMAGE019
Then, the node access set of the current iteration step number is traversed
Figure 972729DEST_PATH_IMAGE020
Each of the non-accessed nodes in (1)
Figure 947638DEST_PATH_IMAGE021
Querying from the holistic knowledge graph
Figure 364189DEST_PATH_IMAGE021
One-step neighborhood node set of
Figure 715536DEST_PATH_IMAGE022
；
Extracting the current patient potential disease development track from the general knowledge map through an attention capture module, wherein the method comprises the following steps: from
Figure 333599DEST_PATH_IMAGE021
Corresponding neighborhood node set
Figure 744989DEST_PATH_IMAGE022
The sampled attention value in the attention value set of all nodes in (1) is the largestkUsing each node as inference node set of current iteration step number
Figure 386185DEST_PATH_IMAGE023
The non-visited node traversed this time
Figure 72382DEST_PATH_IMAGE021
Is modified to access and will
Figure 13793DEST_PATH_IMAGE023
Is absent in
Figure 330505DEST_PATH_IMAGE024
Node of (2) is brought into
Figure 724577DEST_PATH_IMAGE024
In (1), generate
Figure 417727DEST_PATH_IMAGE025
Meanwhile, setting the label of the newly incorporated node as not-accessed;
when in use
Figure 744803DEST_PATH_IMAGE026
At first, query from the general knowledge map
Figure 497995DEST_PATH_IMAGE027
All relations among all nodes in the cognitive map are supplemented into the cognitive map, then the operation of traversing nodes which are not accessed in the node access set is carried out, and inference nodes in the node access set and the cognitive map are updated;
when the number of iteration steps reachesLOr when the inference nodes in the cognitive map inference node set are not increased any more, ending the iteration process to obtain the cognitive map inference node set.
Further, in the subsystem 2, a node state updating mode of the cognitive map is defined by using the variant graph neural network, and the disease and treatment mode of the patient are obtained based on the general knowledge map and the cognitive map, specifically:
reasoning node set of cognitive map obtained in subsystem 1
Figure 113784DEST_PATH_IMAGE028
And reasoning aboutAll the relations of the nodes in the general knowledge graph are brought into the cognitive graph, all the nodes of the cognitive graph are collected and recorded as V,
Figure 610625DEST_PATH_IMAGE029
initial clinical diagnostic data set
Figure 57786DEST_PATH_IMAGE030
，nThe number of data pieces is initiated for the patient,
Figure 716301DEST_PATH_IMAGE031
is as followsiA piece of patient initial data;
setting iteration steps needed by updating node states of cognitive atlasTNodes in cognitive maps
Figure 819386DEST_PATH_IMAGE032
Initial state vector representation of
Figure 854338DEST_PATH_IMAGE033
(ii) a First, the
Figure 421586DEST_PATH_IMAGE034
When step iterates, triplets
Figure 251001DEST_PATH_IMAGE035
Vectorized representation of implied medical knowledge
Figure 841383DEST_PATH_IMAGE036
Wherein
Figure 677096DEST_PATH_IMAGE037
Representing a global vector of a node in an initial clinical diagnosis data set S of a patient;
pairing nodes in cognitive mapsvAggregating medical knowledge contained in all the triples to obtain a triplet aggregation vector
Figure 36533DEST_PATH_IMAGE038
Wherein
Figure 99167DEST_PATH_IMAGE039
For recognizing nodes in a graph spectrum
Figure 176845DEST_PATH_IMAGE040
The number of all triplets in which it resides;
node point
Figure 553599DEST_PATH_IMAGE040
Is/are as follows
Figure 33122DEST_PATH_IMAGE041
State vector of step
Figure 266657DEST_PATH_IMAGE042
Is updated in a manner that
Figure 831631DEST_PATH_IMAGE043
；
Figure 746497DEST_PATH_IMAGE044
And
Figure 80527DEST_PATH_IMAGE045
is a function of two hidden layers.
Further, in the subsystem 2, the state vector of each node of the cognitive map with the updated node state is represented by replacing the global vector of the node in the general knowledge map, all the nodes of the general knowledge map are represented by a matrix X, and the dimension of the matrix X isJ×QWhereinJThe number of nodes in the global knowledge graph,Qa vector dimension for each node; using a two-layer fully-connected network as the final prediction layerF：
Figure 484963DEST_PATH_IMAGE046
The prediction is the most possible disease D of the patient or the recommended treatment mode P given after reasoning, the real disease or the real treatment mode of the patient is marked as target, a sigmoid function is used as an activation function, the error between the prediction and the target is calculated by using a cross entropy loss function, and finally, the cross entropy loss function gradient is propagated by using a random gradient descent algorithm to complete the optimization of the cognitive atlas-based reasoning system.
Furthermore, in the recommendation result display module, existing symptoms and diseases of the patient, treatment modes which may be needed after reasoning and diseases which may be suffered are visually displayed by using the front end of the webpage, and a clinician is supported to input clinical diagnosis and treatment information of the patient during the treatment, so that the diagnosis and treatment path and scheme of the patient are visually displayed.
The beneficial effects of the invention are:
1. according to the technical scheme, the cognitive process of human is referred, and the inference method based on the cognitive map is used for realizing the crossover from perception intelligence to cognitive intelligence, so that the system can truly simulate the diagnosis and treatment thought of a clinician, and an interpretable clinical assistant decision tool with high acceptability is provided for the clinician;
2. the system provided by the technical scheme can be independently used in a clinical scene, provides diagnosis support for doctors and patients, supports real-time interaction between the doctors and the system, and supplements more comprehensive patient information for the system in time, so that more accurate diagnosis suggestion is provided for the patients;
3. aiming at the problem of undifferentiated diseases in the whole department, the technical scheme sets up a personalized diagnosis and treatment scheme for the patient from symptoms, helps the patient to discover the cause of disease as soon as possible and receive targeted treatment, and can realize early screening of dangerous diseases and prompt the patient to transfer to specialized treatment in time.
Drawings
Fig. 1 is a structural diagram of a global patient personalized diagnosis and treatment plan recommendation system based on a cognitive atlas according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the construction of a general knowledge graph according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of constructing a personalized cognitive map of a patient according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a recommended diagnosis path and a result according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those specifically described and will be readily apparent to those of ordinary skill in the art without departing from the spirit of the present invention, and therefore the present invention is not limited to the specific embodiments disclosed below.
The embodiment of the invention provides a cognitive atlas-based individualized diagnosis and treatment scheme recommendation system for general patients, which comprises a data acquisition module, a data preprocessing module, a data analysis reasoning module and a recommendation result display module, as shown in figure 1, and specifically comprises the following steps:
a data acquisition module: the system is used for extracting clinical diagnosis and treatment data of general outpatient service from a medical institution database.
A data preprocessing module: the data acquired by the data acquisition module is preprocessed, and the preprocessing mainly comprises the operations of text data structuring processing, data normalization and the like of the electronic medical record of the outpatient service.
The data analysis reasoning module: the part is the core of the system, firstly a general knowledge graph is constructed, then a cognitive graph of a personalized disease development track of a patient is established by utilizing a designed variant graph neural network and an attention capture module based on information of diseases, symptoms, medicines and the like when the patient is in a treatment and the constructed general knowledge graph, and further personalized diagnosis and treatment scheme recommendation of the patient is given.
A recommendation result display module: the reasoning process of the data analysis reasoning module is displayed visually, and clear personalized disease development and treatment process of the patient are provided.
The following description further provides partial embodiments of implementation of each module of the cognitive-atlas-based general patient personalized diagnosis and treatment scheme recommendation system according to the requirements of the application.
First, data acquisition module
In the embodiment, 22 clinically common symptoms are selected, which are nausea, vomiting, chest pain, lumbago, abdominal pain, hematochezia, dysphagia, gastrointestinal hemorrhage, abdominal distension, fecal incontinence, constipation, diarrhea, urinary incontinence, nocturia, emaciation, weight loss, frequent micturition, urgency, heartburn, dark stool, dysuria and macroscopic hematuria, and all clinical diagnosis and treatment data of a patient from the symptoms to an outpatient clinic visit of a medical institution, including outpatient electronic medical record, personal statistical data, diagnosis data, medication data, operation data, examination result data and the like of the patient, are extracted from a medical institution database by using an SQL command.
Second, data preprocessing module
The information of the chief complaints, the past medical history, the present medical history, the family history and the like of the patients in the clinic are stored in the electronic medical records in the form of texts, and the information of the symptoms, the illness records, the medication records and the like of the patients is extracted from the text data by adopting a method based on rules and a domain dictionary, so that the purpose of text data structuring is achieved.
For the types of the data containing numerical values of the examination results, such as various examination indexes covered by blood routine and urine routine, only the data with positive index results are reserved. According to the technical scheme, the cognitive map method is adopted, and the number of input variables of each piece of data does not need to be fixed, so that missing data interpolation operation is not involved.
A piece of available data is defined as the set of all clinical data for a natural day starting at a certain visit time of the patient. The invention is mainly used for reasoning the disease trend of a patient and giving a treatment suggestion by taking the initial disease and symptoms of the patient as starting points based on the general knowledge graph, so that the patient data which is extracted from a data acquisition module and is periodically sent to a medical institution for prescription, periodic physical examination and periodic review is removed, and the contents of the general knowledge graph which is subsequently constructed are mainly western medicine, so that the patient data of the doctor in the clinic of traditional Chinese medicine is also removed.
Third, data analysis reasoning module
The data analysis reasoning module comprises a general knowledge graph construction module and a reasoning system based on a cognitive graph, and specifically comprises the following steps:
3.1 Whole-discipline knowledge map construction module
The technical scheme provided by the invention is based on the domain knowledge graph, so that the general knowledge graph is required to be constructed before data training. The construction process of the general knowledge graph adopts a mode of combining manual construction and automatic construction, and is shown in figure 2.
(a) In this embodiment, 10 clinicians are temporarily engaged to complete the manual construction of the full-scale knowledge map of the symptoms and disease-related knowledge extracted by the data acquisition module. Specifically, the doctor collects relevant clinical guidelines, expert consensus, professional textbooks, contents on the update website and the like according to symptoms and diseases extracted by the data acquisition module, and manually extracts clinical entities and relations in the collected text contents. In order to standardize the extraction structure form, a text extraction template needs to be designed manually.
(b) The automatic knowledge graph construction part covers relevant clinical terms and relations extracted from the SemMed DB and Chinese ICD-10. The SemMed DB covers rich English triple relations extracted from the medical literature database PubMed, and further calls a translation software API to translate the English triple relations into a Chinese form. The Chinese ICD-10 covers the classification relation and codes of all diseases, and the extraction of relevant disease classifications can fully supplement the hierarchical structure of the constructed general knowledge map.
3.2 inference System based on cognitive maps
The cognitive map is a means for realizing cognitive intelligence, the cognitive map technology simulates the idea of solving problems of human beings, and the two-system theory provided in the field of brain cognitive science is referred to. The cognitive process of the double-system theory thought person comprises two systems: the system 1 is based on an intuitionistic and unconscious thinking system, is a perception process, and the system 2 is explicit and logical, needs consciousness control and is the embodiment of high-level intelligence of human beings. The inference system based on the cognitive map comprises two subsystems, wherein the subsystem 1 is a process for exploring a disease development track on a constructed general knowledge map based on initial clinical diagnosis and treatment data of a patient, and the subsystem 2 is a process for accurately positioning the final state of the patient based on a plurality of potential disease development tracks of the patient explored by the subsystem 1.
As shown in FIG. 3, the constructed global knowledge graph is denoted G, which contains a series of triples
Figure 537233DEST_PATH_IMAGE047
Each triplet representing a piece of medical knowledge and each triplet being represented by a relationship
Figure 990211DEST_PATH_IMAGE048
And two entities
Figure 444326DEST_PATH_IMAGE049
Composition of, whereinrBy head entityhPointing to tail entitiest，
Figure 222926DEST_PATH_IMAGE050
The set of all the entities of the general knowledge graph G is shown, each entity in the triplets serves as a node in the general knowledge graph G, and R is the set of all the relationships of the general knowledge graph G. The entity types include: symptoms, diseases, examination means, treatment methods, administration, examination results, and the like. These triplets contain relationships between symptoms and disease, disease and means of examination, disease and means of treatment, and the like. The technical scheme aims to obtain the disease D and the treatment mode P of a patient through reasoning based on an initial clinical diagnosis and treatment data set S of the patient and give an intermediate reasoning process, wherein the initial clinical diagnosis and treatment data set S
Figure 559230DEST_PATH_IMAGE051
The initial clinical diagnosis and treatment data set is sorted according to the time sequence of each symptom, medical history and medicine data generation when the patient initially visits the clinic,nthe number of pieces of data is initiated for the patient,
Figure 815899DEST_PATH_IMAGE052
is as followsiPatient initial data.
Patient-customized cognitive profiles
Figure 124520DEST_PATH_IMAGE053
Initially only from the patient in the initial clinical data set SnFormed from initial patient datanAnd each node is formed.
The subsystem 1 firstly carries out global vectorization on the general knowledge graph to realize semantic intercommunication of graph global nodes, and then the subsystem is based on the cognitive graph
Figure 870759DEST_PATH_IMAGE053
And the global vectorized general knowledge graph iteratively captures a potential cognitive graph inference node set
Figure 632042DEST_PATH_IMAGE054
Cognitive atlas inference node set
Figure 692402DEST_PATH_IMAGE054
The node in (1) may be the disease of the patient, the examination to be made or the medicine to be taken;
the subsystem 2 infers the cognitive map to a node set
Figure 876038DEST_PATH_IMAGE055
The nodes in the cognitive map are gradually supplemented to the cognitive map
Figure 996441DEST_PATH_IMAGE053
Then, defining a node state updating mode of the cognitive map, and finally obtaining the disease D and the treatment mode P of the patient based on the general knowledge map and the cognitive map.
In one embodiment, the implementation of the subsystem 1 is specifically: the global vectorization of the general knowledge Graph is realized by a variant network (GNN) based on Graph Neural Network (GNN), and then a traditional attention thought is simulated to design an attention capturing module of the potential disease development track of a patient, wherein the module is based on an initial attention capturing moduleClinical diagnostic data set
Figure 307336DEST_PATH_IMAGE056
And iteratively capturing inference nodes required in the inference process of the cognitive map from the globally vectorized general knowledge map.
First, the global vectorization process of the holistic knowledge graph is introduced. For each node of the holistic knowledge graphhDefining a learnable embedded vector representation
Figure 171387DEST_PATH_IMAGE057
For each relationshiprAlso defined is a learnable embedded vector representation
Figure 923442DEST_PATH_IMAGE058
. The global vectorization process of the general knowledge graph needs to be carried outMIteration of (MHuman setting, in this embodimentM = 3), the second
Figure 480326DEST_PATH_IMAGE059
When the steps are iterated, the node state vector is expressed as
Figure 12938DEST_PATH_IMAGE060
To do so
Figure 680680DEST_PATH_IMAGE061
. Each triplet
Figure 287242DEST_PATH_IMAGE062
Represents a piece of medical knowledge, inmCan be vectorized into
Figure 15026DEST_PATH_IMAGE063
. In the global vectorization process of the general knowledge graph, the nodes are relatedhRelated medical knowledge
Figure 34935DEST_PATH_IMAGE064
The vectorization of (c) is represented as follows:
Figure 240788DEST_PATH_IMAGE065
wherein
Figure 233015DEST_PATH_IMAGE066
,
Figure 866122DEST_PATH_IMAGE067
Are respectively the firstmStep iteration time nodeh，tA node state vector of; the vectorized representation of the medical knowledge relates to a vector representation of head and tail entities and relationships in the medical knowledge;
node pointhThe state updating in the iteration process is related to the vector representation of the state updating and the vector representation of all triples, namely the neighborhood, where the state updating is located, and the nodes are located in the general knowledge graphhAggregating medical knowledge contained in all the triples to obtain a triplet aggregation vector
Figure 311009DEST_PATH_IMAGE068
It can be expressed as:
Figure 117291DEST_PATH_IMAGE069
wherein
Figure 229604DEST_PATH_IMAGE070
Is a node in the general knowledge graphhThe number of all triplets in which they are located;
Figure 768033DEST_PATH_IMAGE071
calculating by aggregating vector information of related triples and dividing by a neighborhood number root-mean-square;
final nodehOf the next step state vector
Figure 965796DEST_PATH_IMAGE072
The updating method comprises the following steps:
Figure 310189DEST_PATH_IMAGE073
Figure 277008DEST_PATH_IMAGE074
and node initial state vector
Figure 983409DEST_PATH_IMAGE075
State vector of current iteration step
Figure 668468DEST_PATH_IMAGE076
And a triplet aggregation vector composed of the current node and its neighborhood nodes
Figure 19815DEST_PATH_IMAGE077
In connection with this, this also expresses the process of neighborhood medical knowledge to deliver information to the current node;
the above
Figure 637878DEST_PATH_IMAGE078
、
Figure 49268DEST_PATH_IMAGE079
And respectively representing two hidden layer functions, wherein the input is the splicing of all vectors, and the output is consistent with the dimension of the input single vector. In addition, the mode of residual connection is adopted for carrying out the nodehTo improve the response withmMay have information loss, and finally selects the output result of the last step
Figure 690465DEST_PATH_IMAGE080
As nodeshThe global vector representation of (a) is input into the next step.
A method for capturing inference nodes required in the inference process of the cognitive map from the global vectorized general knowledge map is described below.
Cognitive map
Figure 579923DEST_PATH_IMAGE081
Initially by the patientInitial clinical diagnostic data set
Figure 318072DEST_PATH_IMAGE082
Forming, defining a node access set
Figure 634784DEST_PATH_IMAGE083
Initial node Access Collection
Figure 763277DEST_PATH_IMAGE084
And will be
Figure 456427DEST_PATH_IMAGE085
Each node in the set is marked as no-visited, and the maximum iteration step is set to beL（LHuman setting, in this embodimentL = 6). When the number of iteration steps is
Figure 49082DEST_PATH_IMAGE086
Then, the node access set of the current iteration step number is traversed
Figure 802274DEST_PATH_IMAGE087
Each of the non-visited no-visited nodes in (1)
Figure 683643DEST_PATH_IMAGE088
From the global knowledge map
Figure 914904DEST_PATH_IMAGE089
One-step neighborhood node set
Figure 299749DEST_PATH_IMAGE090
. Although we are only concerned with
Figure 223843DEST_PATH_IMAGE089
However, as the number of iteration steps increases, the neighborhood node set of the traversed nodes gradually expands, and in order to reduce the amount of computation, the number of traversed nodes needs to be controlled.
The present embodiment proposes an attention capture module for capturing attention from the outsideExtracting the current potential disease development track of the patient from the department knowledge map, wherein the process can control the expansion speed of the cognitive map, and the implementation process of the attention capturing module is as follows: from
Figure 389245DEST_PATH_IMAGE091
Middle sampling with the largest attention valuekUsing each node as inference node set of current iteration step number
Figure 424197DEST_PATH_IMAGE092
In which
Figure 663548DEST_PATH_IMAGE091
Is composed of
Figure 555281DEST_PATH_IMAGE093
Corresponding neighborhood node set
Figure 408312DEST_PATH_IMAGE094
The attention values of all the nodes in the traversal are collected, and the non-visited nodes of the traversal are collected
Figure 981375DEST_PATH_IMAGE093
Will all access the visited instead and will
Figure 340813DEST_PATH_IMAGE095
Is absent in
Figure 606709DEST_PATH_IMAGE096
Node of (D) is incorporated into
Figure 481124DEST_PATH_IMAGE096
In (1), generate
Figure 123458DEST_PATH_IMAGE097
The labels of these newly incorporated nodes are set to no-visited at the same time. In addition, when
Figure 337402DEST_PATH_IMAGE098
When it comes, it is first necessary to extract the knowledge-graph from the whole familyQuery in
Figure 774199DEST_PATH_IMAGE099
All relations among all nodes in the cognitive map are supplemented to the cognitive map
Figure 135910DEST_PATH_IMAGE100
And then, performing the operation of traversing nodes which are not accessed in the node access set, and updating inference nodes in the node access set and the cognitive map. When the number of iteration steps reachesLOr when the inference nodes in the cognitive map inference node set are not increased any more, the iteration process is ended, and the obtained cognitive map inference node set is recorded as
Figure 50777DEST_PATH_IMAGE101
。
In one embodiment, the implementation of the subsystem 2 is specifically: defining a node state updating mode of the cognitive map by using a variant network of the graph neural network GNN, and finally obtaining the disease of the patient based on the general knowledge map and the cognitive map inference
Figure 384806DEST_PATH_IMAGE102
And the mode of treatment
Figure 992505DEST_PATH_IMAGE103
。
Firstly, a cognitive map inference node set finally obtained in a subsystem 1 is subjected to node set inference
Figure 841512DEST_PATH_IMAGE104
And all relations of the reasoning nodes in the general knowledge graph G are included in the cognitive graph
Figure 560070DEST_PATH_IMAGE100
In (1), all node sets of the cognitive map are marked as V, wherein
Figure 14185DEST_PATH_IMAGE105
，
Figure 527206DEST_PATH_IMAGE106
Then defining the number of iteration steps needed by updating the node state of the cognitive mapT（THuman setting, in this embodimentT = 6). Cognitive map
Figure 66771DEST_PATH_IMAGE100
Node in
Figure 589020DEST_PATH_IMAGE107
Initial vector representation of
Figure 694379DEST_PATH_IMAGE108
. Nodes in cognitive mapsvIn the first place
Figure 378301DEST_PATH_IMAGE109
The updating of the node state of the step is related to the node state of the step at the 0 th step and the t th step and the vector representation of the triad where the node state of the step is located in the cognitive map. Triplet with iteration step number t
Figure 136654DEST_PATH_IMAGE110
The medical knowledge of implications can be vectorized into
Figure 259331DEST_PATH_IMAGE111
The calculation method is as follows:
Figure 422459DEST_PATH_IMAGE112
wherein
Figure 277282DEST_PATH_IMAGE113
The global vector representing the nodes in the initial clinical diagnosis and treatment data set S of the patient is represented by the subsystem 1, and because the information such as symptoms, diseases or medication and the like represented by the nodes in the S is the clinical diagnosis and treatment data really generated by the patient and is all starting points of reasoning, the information is included in the updated calculation of each step;
pairing nodes in cognitive mapsvAll triplets of medicineThe knowledge is aggregated to obtain a triple aggregation vector
Figure 791440DEST_PATH_IMAGE114
It can be expressed as:
Figure 655491DEST_PATH_IMAGE115
wherein
Figure 469863DEST_PATH_IMAGE116
For recognizing nodes in a graph spectrum
Figure 26747DEST_PATH_IMAGE117
The number of all triplets in which it resides; similar to the subsystem 1, the vector information of the relevant triples is aggregated and then divided by the root-mean-square of the number of neighborhoods to calculate;
node point
Figure 762621DEST_PATH_IMAGE117
Is/are as follows
Figure 899205DEST_PATH_IMAGE118
The state vector update mode of the steps is as follows:
Figure 833663DEST_PATH_IMAGE119
Figure 561447DEST_PATH_IMAGE120
also with nodes
Figure 253460DEST_PATH_IMAGE121
Step 0 state vector of
Figure 724892DEST_PATH_IMAGE122
Node, node
Figure 513857DEST_PATH_IMAGE123
The t-th state vector
Figure 146963DEST_PATH_IMAGE124
Node, node
Figure 857431DEST_PATH_IMAGE123
All triplet aggregation vectors formed by neighborhood nodes of cognitive map
Figure 601396DEST_PATH_IMAGE125
And a global vector representation of the nodes in the initial clinical data set S of the patient.
As described above
Figure 448129DEST_PATH_IMAGE126
、
Figure 314454DEST_PATH_IMAGE127
The node state is updated by adopting a residual error connection mode.
Due to cognitive mapping
Figure 246638DEST_PATH_IMAGE100
All the nodes in the cognitive map are from the general knowledge map G, and the cognitive map after the updating is finished at the moment is used
Figure 791364DEST_PATH_IMAGE100
The state vector of each node replaces the global vector representation in the general knowledge graph G, so that all nodes of the whole general knowledge graph G can be represented by a matrix X with the dimension ofJ×QWhereinJThe number of nodes in the global knowledge graph,Qas a vector dimension for each node. Using a two-layer fully-connected network as the final prediction layerF：
Figure 758183DEST_PATH_IMAGE128
The prediction is the most possible disease D or the recommended treatment mode P of the patient after inference is carried out, the actual disease or the actual treatment mode of the patient is target, a sigmoid function is used as an activation function, the error between the predicted value prediction and the actual value target is calculated by using a cross entropy loss function, and finally, a random gradient descent (SGD) algorithm is used for propagating the cross entropy loss function gradient to complete the optimization of the inference system based on the cognitive atlas.
Fourth, recommend the explicit module of result
The biggest advantage of the general patient personalized diagnosis and treatment scheme recommendation system based on the cognitive map is interpretability, diagnosis and treatment ideas of clinicians can be simulated, and a clear reasoning process is given, so that the recommendation result display module mainly utilizes the front end of a webpage to visually display existing symptoms and diseases of a patient, treatment modes possibly required after reasoning and diseases possibly suffered by the patient. The webpage supports the clinician to input clinical diagnosis and treatment information when the patient visits, and after clicking analysis, the diagnosis and treatment path and scheme of the patient can be visually displayed. Fig. 4 shows the recommended diagnosis and treatment routes and results of an embodiment, in fig. 4, the dotted line box represents clinical diagnosis and treatment information of a patient at the time of treatment, the patient has a history of hypertension, peptic ulcer, gastritis and the like, and takes the gastric drug domperidone all the time, and the appetite is reduced recently, so that the patient can see the treatment, and then the system prompts the examination to be performed, and diseases such as irritable bowel syndrome, colorectal cancer and benign tumor are gradually eliminated, so that diagnosis and treatment suggestions that the patient needs to be subjected to gastroscopic resection are finally given, wherein the early gastric cancer is the patient.
The foregoing is only a preferred embodiment of the present invention, and although the present invention has been disclosed in the preferred embodiments, it is not intended to limit the present invention. Those skilled in the art can make numerous possible variations and modifications to the present teachings, or modify equivalent embodiments to equivalent variations, without departing from the scope of the present teachings, using the methods and techniques disclosed above. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention are within the scope of the technical solution of the present invention, unless the technical essence of the present invention is not departed from the content of the technical solution of the present invention.

Claims (7)

1. A general patient personalized diagnosis and treatment scheme recommendation system based on a cognitive map is characterized by comprising:
a data acquisition module: the system is used for extracting clinical diagnosis and treatment data of the general outpatient clinic from a medical institution database;
a data preprocessing module: carrying out data preprocessing operation on the data acquired by the data acquisition module, including the structured processing and data normalization of the text data of the electronic medical record of the outpatient department;
the data analysis reasoning module: constructing a general knowledge map, establishing a cognitive map of a personalized disease development track of a patient by using a variant map neural network and an attention capture module based on diseases, symptoms and medication information of the patient in the clinic and the constructed general knowledge map, and recommending a personalized diagnosis and treatment scheme of the patient;
the data analysis reasoning module comprises a general knowledge graph construction module and a reasoning system based on a cognitive graph; the cognitive map-based reasoning system comprises two subsystems: the subsystem 1 is used for carrying out disease development track exploration on the constructed full-knowledge map based on the initial clinical diagnosis and treatment data set of the patient; the subsystem 2 is used for accurately positioning the final state of the patient based on a plurality of potential disease development tracks of the patient explored by the subsystem 1;
the subsystem 1 comprises: carrying out global vectorization on the general knowledge graph; the individualized cognitive map of the patient is initially formed by nodes formed by a plurality of initial patient data in an initial clinical diagnosis and treatment data set of the patient, and a potential cognitive map inference node set is iteratively captured on the basis of the cognitive map and a global vectorized general knowledge map; the method specifically comprises the following steps:
defining node access sets
Figure DEST_PATH_IMAGE001
Initial node Access Collection
Figure DEST_PATH_IMAGE002
S is the initial clinical data set of the patient, will
Figure DEST_PATH_IMAGE003
Each node in the system is marked as not-accessed, and the maximum iteration step number is setLWhen the number of iteration steps is
Figure DEST_PATH_IMAGE004
Then, the node access set of the current iteration step number is traversed
Figure DEST_PATH_IMAGE005
Each of the non-accessed nodes in (1)
Figure DEST_PATH_IMAGE006
Querying from the holistic knowledge graph
Figure 500439DEST_PATH_IMAGE006
One-step neighborhood node set of
Figure DEST_PATH_IMAGE007
；
Extracting the current potential disease development track of the patient from the general knowledge map through an attention capturing module, wherein the track comprises the following steps: from
Figure DEST_PATH_IMAGE008
Corresponding neighborhood node set
Figure 729164DEST_PATH_IMAGE007
The sampled attention value in the attention value set of all the nodes in (1) is the largestkUsing each node as inference node set of current iteration step number
Figure DEST_PATH_IMAGE009
The non-visited node traversed this time
Figure 243322DEST_PATH_IMAGE008
Is modified to access and will
Figure 372952DEST_PATH_IMAGE009
Is absent in
Figure DEST_PATH_IMAGE010
Node of (2) is brought into
Figure 46379DEST_PATH_IMAGE010
In (1), generate
Figure DEST_PATH_IMAGE011
Meanwhile, setting the label of the newly-incorporated node as not-accessed;
when the temperature is higher than the set temperature
Figure DEST_PATH_IMAGE012
At first, query from the general knowledge map
Figure DEST_PATH_IMAGE013
All relations among all nodes in the cognitive map are supplemented into the cognitive map, then the operation of traversing nodes which are not accessed in the node access set is carried out, and inference nodes in the node access set and the cognitive map are updated;
when the number of iteration steps reachesLOr when the inference nodes in the cognitive map inference node set are not increased any more, ending the iteration process to obtain a cognitive map inference node set;
the subsystem 2 comprises: gradually supplementing the nodes in the cognitive map inference node set into the cognitive map, defining a node state updating mode of the cognitive map, and inferring and obtaining diseases and treatment modes of the patient based on the general knowledge map and the cognitive map;
a recommendation result display module: and the reasoning process of the data analysis reasoning module is visually displayed, and the personalized disease development and treatment process of the patient is provided.
2. The system for recommending the personalized diagnosis and treatment plan of the general patient based on the cognitive atlas as claimed in claim 1, wherein in the data preprocessing module, the method based on rules and domain dictionary is adopted to extract the symptoms, the illness records and the medication records of the patient from the text data of the electronic medical record in the outpatient service, so as to realize the text data structuring;
only data with positive index results are reserved according to the data types of the examination results containing numerical values; removing patient data which is periodically prescribed, periodically checked and periodically reviewed by a medical institution; remove the data of the patients in the clinic of traditional Chinese medicine.
3. The system for recommending a personalized diagnosis and treatment plan for a general patient based on a cognitive atlas as claimed in claim 1, wherein in the data analysis reasoning module, the general knowledge atlas is constructed manually and automatically; the clinical doctor completes the manual construction part of the general knowledge map of the symptom and disease related knowledge extracted by the data acquisition module; the automatic construction part of the general knowledge map covers relevant clinical terms and relations extracted from the SemMed DB and Chinese ICD-10.
4. The system for recommending the personalized diagnosis and treatment plan of the general patient based on the cognitive atlas as claimed in claim 1, wherein in the subsystem 1, global vectorization of the general knowledge atlas is realized based on a variogram neural network, specifically:
the global knowledge graph comprises a series of triplets
Figure DEST_PATH_IMAGE014
Each triplet consisting of a relationshiprAnd two entities
Figure DEST_PATH_IMAGE015
Each entity is used as a node in a general knowledge graph, and each triple represents a piece of medical knowledge;
for each node of the global knowledge graphhDefine a learnableEmbedded vector representation of
Figure DEST_PATH_IMAGE016
For each relationshiprDefining a learnable embedded vector representation
Figure DEST_PATH_IMAGE017
(ii) a Setting global vectorization process iteration stepsMOf 1 at
Figure DEST_PATH_IMAGE018
While iterating, the nodehIs represented as a state vector of
Figure DEST_PATH_IMAGE019
，
Figure DEST_PATH_IMAGE020
And nodehRelated medical knowledge
Figure 150732DEST_PATH_IMAGE014
Vectorized representation of
Figure DEST_PATH_IMAGE021
；
Pairing nodes in a holistic knowledge graphhAggregating medical knowledge contained in all the triples to obtain a triplet aggregation vector
Figure DEST_PATH_IMAGE022
In which
Figure DEST_PATH_IMAGE023
Is a node in the general knowledge graphhThe number of all triplets in which they are located;
node pointhOf the next step state vector
Figure DEST_PATH_IMAGE024
In an update manner of
Figure DEST_PATH_IMAGE025
；
Figure DEST_PATH_IMAGE026
And
Figure DEST_PATH_IMAGE027
the input is the splicing of all vectors for two hidden layer functions, and the output is consistent with the dimension of the input single vector;
the output result of the last step
Figure DEST_PATH_IMAGE028
As nodeshIs represented by a global vector.
5. The system for recommending a personalized diagnosis and treatment plan for a general patient based on a cognitive map as claimed in claim 4, wherein in the subsystem 2, a node state updating mode of the cognitive map is defined by using a variant graph neural network, and a disease and treatment mode of the patient is obtained based on the general knowledge map and the cognitive map, specifically:
reasoning node set of cognitive map obtained in subsystem 1
Figure DEST_PATH_IMAGE029
And all the relationships of the reasoning nodes in the general knowledge graph are brought into the cognitive graph, all the nodes of the cognitive graph are collected and recorded as V,
Figure DEST_PATH_IMAGE030
initial clinical data set
Figure DEST_PATH_IMAGE031
，nThe number of data pieces is initiated for the patient,
Figure DEST_PATH_IMAGE032
is as followsiPatients with bandingInitial data;
setting iteration steps required by updating node states of cognitive mapsTNodes in cognitive maps
Figure DEST_PATH_IMAGE033
Initial state vector representation of
Figure DEST_PATH_IMAGE034
(ii) a First, the
Figure DEST_PATH_IMAGE035
When step iterates, triplets
Figure DEST_PATH_IMAGE036
Vectorized representation of implied medical knowledge
Figure DEST_PATH_IMAGE037
Wherein
Figure DEST_PATH_IMAGE038
Representing the global vector of the node in the initial clinical diagnosis and treatment data set S of the patient;
pairing nodes in cognitive mapsvAggregating medical knowledge contained in all the triples to obtain a triplet aggregation vector
Figure DEST_PATH_IMAGE039
Wherein
Figure DEST_PATH_IMAGE040
For recognizing nodes in a graph spectrum
Figure DEST_PATH_IMAGE041
The number of all triplets in which they are located;
node point
Figure 932612DEST_PATH_IMAGE041
Is/are as follows
Figure DEST_PATH_IMAGE042
State vector of step
Figure DEST_PATH_IMAGE043
In an update manner of
Figure DEST_PATH_IMAGE044
；
Figure DEST_PATH_IMAGE045
And
Figure DEST_PATH_IMAGE046
is a function of two hidden layers.
6. The cognitive-atlas-based individualized diagnosis and treatment scheme recommendation system for general patients according to claim 5, wherein in the subsystem 2, the state vector of each node of the cognitive atlas with the updated node state replaces the global vector representation of the node in the general knowledge atlas, all the nodes of the general knowledge atlas are represented by a matrix X, and the dimension of the matrix X isJ×QIn whichJThe number of nodes in the global knowledge graph,Qa vector dimension for each node; adopting a two-layer fully-connected network as a final prediction layerF：
Figure DEST_PATH_IMAGE047
The prediction is the most possible disease D of the patient or the recommended treatment mode P given after reasoning, the real disease or the real treatment mode of the patient is marked as target, a sigmoid function is used as an activation function, the error between the prediction and the target is calculated by using a cross entropy loss function, and finally, the cross entropy loss function gradient is propagated by using a random gradient descent algorithm to complete the optimization of the cognitive atlas-based reasoning system.
7. The system for recommending a personalized diagnosis and treatment plan for a general patient based on a cognitive atlas according to any one of claims 1-6, wherein the recommendation result display module utilizes a front end of a webpage to visually display existing symptoms and diseases of the patient, treatment modes which may be needed after reasoning and diseases which may be suffered by the patient, and supports a clinician to input clinical diagnosis and treatment information of the patient during the visit so as to visually display the diagnosis and treatment path and plan of the patient.
CN202210403938.2A 2022-04-18 2022-04-18 Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients Active CN114496234B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202210403938.2A CN114496234B (en) 2022-04-18 2022-04-18 Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients
PCT/CN2023/088590 WO2023202508A1 (en) 2022-04-18 2023-04-17 Cognitive graph-based general practice patient personalized diagnosis and treatment scheme recommendation system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210403938.2A CN114496234B (en) 2022-04-18 2022-04-18 Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients

Publications (2)

Publication Number Publication Date
CN114496234A CN114496234A (en) 2022-05-13
CN114496234B true CN114496234B (en) 2022-07-19

Family

ID=81489383

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210403938.2A Active CN114496234B (en) 2022-04-18 2022-04-18 Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients

Country Status (2)

Country Link
CN (1) CN114496234B (en)
WO (1) WO2023202508A1 (en)

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114496234B (en) * 2022-04-18 2022-07-19 浙江大学 Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients
CN114969557B (en) * 2022-07-29 2022-11-08 之江实验室 Propaganda and education pushing method and system based on multi-source information fusion
CN115359870B (en) * 2022-10-20 2023-03-24 之江实验室 Disease diagnosis and treatment process abnormity identification system based on hierarchical graph neural network
CN115658877B (en) * 2022-12-27 2023-03-21 神州医疗科技股份有限公司 Medicine recommendation method and device based on reinforcement learning, electronic equipment and medium
CN115862848B (en) * 2023-02-15 2023-05-30 之江实验室 Disease prediction system and device based on clinical data screening and medical knowledge graph
CN115858876B (en) * 2023-02-21 2023-05-26 杭州华卓信息科技有限公司 Follow-up content intelligent pushing method and system based on disease knowledge graph
CN116187448B (en) * 2023-04-25 2023-08-01 之江实验室 Information display method and device, storage medium and electronic equipment
CN116682553B (en) * 2023-08-02 2023-11-03 浙江大学 Diagnosis recommendation system integrating knowledge and patient representation
CN117012374B (en) * 2023-10-07 2024-01-26 之江实验室 Medical follow-up system and method integrating event map and deep reinforcement learning
CN117012375B (en) * 2023-10-07 2024-03-26 之江实验室 Clinical decision support method and system based on patient topological feature similarity
CN117153378B (en) * 2023-10-31 2024-03-01 北京博晖创新生物技术集团股份有限公司 Diagnosis guiding method and device, electronic equipment and storage medium
CN117831703B (en) * 2024-01-02 2024-08-09 复旦大学 Method for recommending care plan of disabled old people based on graph neural network and deep learning
CN117524405B (en) * 2024-01-05 2024-03-26 长春中医药大学 Cloud computing-based gynecological nursing method intelligent selection system
CN117649949B (en) * 2024-01-29 2024-04-30 浙江大学 Clinical thinking data generation system and method based on reinforcement learning
CN117973947B (en) * 2024-04-01 2024-06-14 国网山东省电力公司宁津县供电公司 Standardized acceptance checking method and system for power distribution network engineering construction process
CN118016316B (en) * 2024-04-10 2024-06-04 健数(长春)科技有限公司 Disease screening rate improving method and system by combining knowledge graph with blood routine test data
CN118116584A (en) * 2024-04-23 2024-05-31 鼎泰(南京)临床医学研究有限公司 Big data-based adjustable medical auxiliary diagnosis system and method
CN118674055A (en) * 2024-04-26 2024-09-20 支付宝(杭州)信息技术有限公司 LLM model reasoning method based on medical knowledge graph and related equipment

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1999004043A1 (en) * 1997-07-14 1999-01-28 Abbott Laboratories Telemedicine
US6687685B1 (en) * 2000-04-07 2004-02-03 Dr. Red Duke, Inc. Automated medical decision making utilizing bayesian network knowledge domain modeling
CN104573350A (en) * 2014-12-26 2015-04-29 深圳市前海安测信息技术有限公司 System and method for general practitioner auxiliary diagnosis and therapy based on network hospital
CN108461151A (en) * 2017-12-15 2018-08-28 北京大学深圳研究生院 A kind of the logic Enhancement Method and device of knowledge mapping
CN109920540A (en) * 2019-03-14 2019-06-21 宁波中云创科信息技术有限公司 Construction method, device and the computer equipment of assisting in diagnosis and treatment decision system
KR20190135908A (en) * 2019-02-01 2019-12-09 (주)제이엘케이인스펙션 Artificial intelligence based dementia diagnosing method and apparatus
CN111767410A (en) * 2020-06-30 2020-10-13 平安国际智慧城市科技股份有限公司 Construction method, device, equipment and storage medium of clinical medical knowledge map
CN112102937A (en) * 2020-11-13 2020-12-18 之江实验室 Patient data visualization method and system for chronic disease assistant decision making
CN112463987A (en) * 2020-12-09 2021-03-09 中国园林博物馆北京筹备办公室 Chinese classical garden knowledge graph completion and cognitive reasoning method
WO2021189971A1 (en) * 2020-10-26 2021-09-30 平安科技（深圳）有限公司 Medical plan recommendation system and method based on knowledge graph representation learning
CN113779220A (en) * 2021-09-13 2021-12-10 内蒙古工业大学 Mongolian multi-hop question-answering method based on three-channel cognitive map and graph attention network
CN113990495A (en) * 2021-12-27 2022-01-28 之江实验室 Disease diagnosis prediction system based on graph neural network

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050108052A1 (en) * 2003-11-03 2005-05-19 Omaboe Nortey J. Proces for diagnosic system and method applying artificial intelligence techniques to a patient medical record and that combines customer relationship management (CRM) and enterprise resource planning (ERP) software in a revolutionary way to provide a unique-and uniquely powerful and easy-to-use-tool to manage veterinary or human medical clinics and hospitals
US20210169355A1 (en) * 2019-12-06 2021-06-10 Savan Patel Systems & Methods for Vascular Disease Prediction, Indication, or Diagnosis
CN112037912B (en) * 2020-09-09 2023-07-11 平安科技（深圳）有限公司 Triage model training method, device and equipment based on medical knowledge graph
WO2022072785A1 (en) * 2020-10-01 2022-04-07 University Of Massachusetts A neural graph model for automated clinical assessment generation
CN113010663A (en) * 2021-04-26 2021-06-22 东华大学 Adaptive reasoning question-answering method and system based on industrial cognitive map
CN113889259A (en) * 2021-09-06 2022-01-04 浙江工业大学 Automatic diagnosis dialogue system under assistance of knowledge graph
CN113808693A (en) * 2021-09-10 2021-12-17 浙江科技学院 Medicine recommendation method based on graph neural network and attention mechanism
CN113871003B (en) * 2021-12-01 2022-04-08 浙江大学 Disease auxiliary differential diagnosis system based on causal medical knowledge graph
CN114496234B (en) * 2022-04-18 2022-07-19 浙江大学 Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1999004043A1 (en) * 1997-07-14 1999-01-28 Abbott Laboratories Telemedicine
US6687685B1 (en) * 2000-04-07 2004-02-03 Dr. Red Duke, Inc. Automated medical decision making utilizing bayesian network knowledge domain modeling
CN104573350A (en) * 2014-12-26 2015-04-29 深圳市前海安测信息技术有限公司 System and method for general practitioner auxiliary diagnosis and therapy based on network hospital
CN108461151A (en) * 2017-12-15 2018-08-28 北京大学深圳研究生院 A kind of the logic Enhancement Method and device of knowledge mapping
KR20190135908A (en) * 2019-02-01 2019-12-09 (주)제이엘케이인스펙션 Artificial intelligence based dementia diagnosing method and apparatus
CN109920540A (en) * 2019-03-14 2019-06-21 宁波中云创科信息技术有限公司 Construction method, device and the computer equipment of assisting in diagnosis and treatment decision system
CN111767410A (en) * 2020-06-30 2020-10-13 平安国际智慧城市科技股份有限公司 Construction method, device, equipment and storage medium of clinical medical knowledge map
WO2021189971A1 (en) * 2020-10-26 2021-09-30 平安科技（深圳）有限公司 Medical plan recommendation system and method based on knowledge graph representation learning
CN112102937A (en) * 2020-11-13 2020-12-18 之江实验室 Patient data visualization method and system for chronic disease assistant decision making
CN112463987A (en) * 2020-12-09 2021-03-09 中国园林博物馆北京筹备办公室 Chinese classical garden knowledge graph completion and cognitive reasoning method
CN113779220A (en) * 2021-09-13 2021-12-10 内蒙古工业大学 Mongolian multi-hop question-answering method based on three-channel cognitive map and graph attention network
CN113990495A (en) * 2021-12-27 2022-01-28 之江实验室 Disease diagnosis prediction system based on graph neural network

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
A multicenter random forest model for effective prognosis prediction in collaborative clinical research network;Li,J ET AL;《ARTIFICIAL INTELLIGENCE IN MEDICINE》;20200407;全文 *
A novel lifelong machine learning-based method to eliminate calibration drift in clinical prediction models.;Chi, Shengqiang ET AL;《Artificial intelligence in medicine》;20220305;全文 *
Multimodality Attention-Guided 3-D Detection of Nonsmall Cell Lung Cancer in 18F-FDG PET/CT Images;Ling Chen ET AL;《IEEE Transactions on Radiation and Plasma Medical Sciences》;20220409;第6卷(第4期);全文 *
基于动态图神经网络的会话式机器阅读研究;刘啸等;《集成技术》;20220331;第11卷(第2期);全文 *
基于认知图谱的智能问答系统推理模型研究;袁满等;《吉林大学学报》;20210930;第39卷(第5期);全文 *

Also Published As

Publication number Publication date
CN114496234A (en) 2022-05-13
WO2023202508A1 (en) 2023-10-26

Similar Documents

Publication Publication Date Title
CN114496234B (en) Cognitive-atlas-based personalized diagnosis and treatment scheme recommendation system for general patients
Sullivan Understanding from machine learning models
Yu et al. Clinical big data and deep learning: Applications, challenges, and future outlooks
Johnson et al. Artificial intelligence in cardiology
Arocha et al. Identifying reasoning strategies in medical decision making: a methodological guide
CN113421652B (en) Method for analyzing medical data, method for training model and analyzer
Lee et al. Machine learning in relation to emergency medicine clinical and operational scenarios: an overview
CN113015977A (en) Deep learning based diagnosis and referral of diseases and conditions using natural language processing
CN107247881A (en) A kind of multi-modal intelligent analysis method and system
CN111492437A (en) Method and system for supporting medical decision
WO2023078025A1 (en) Task decomposition strategy-based auxiliary differential diagnosis system for fever of unknown origin
Moein Medical diagnosis using artificial neural networks
CN110880362A (en) Large-scale medical data knowledge mining and treatment scheme recommending system
Fernandez-Blanco et al. EEG signal processing with separable convolutional neural network for automatic scoring of sleeping stage
CN112489769A (en) Intelligent traditional Chinese medicine diagnosis and medicine recommendation system for chronic diseases based on deep neural network
CN109360658A (en) A kind of the disease pattern method for digging and device of word-based vector model
CN117316466B (en) Clinical decision method, system and equipment based on knowledge graph and natural language processing technology
Kumar et al. Deep-learning-enabled multimodal data fusion for lung disease classification
CN117854748A (en) Knowledge graph and generation-based large model inquiry method and system
CN117672450A (en) Personalized medicine recommendation method and system based on knowledge graph
JP7365747B1 (en) Disease treatment process abnormality identification system based on hierarchical neural network
Yadav et al. Machine learning in expert systems for disease diagnostics in human healthcare
Karthikeyan et al. Smart COVIDNet: designing an IoT-based COVID-19 disease prediction framework using attentive and adaptive-derived ensemble deep learning
Al Sukar et al. Identifying a drug addict person using artificial neural networks
Strobel et al. Healthcare in the Era of Digital twins: towards a Domain-Specific Taxonomy.

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant