CN108198621A - A kind of database data synthesis dicision of diagnosis and treatment method based on neural network - Google Patents
A kind of database data synthesis dicision of diagnosis and treatment method based on neural network Download PDFInfo
- Publication number
- CN108198621A CN108198621A CN201810047166.7A CN201810047166A CN108198621A CN 108198621 A CN108198621 A CN 108198621A CN 201810047166 A CN201810047166 A CN 201810047166A CN 108198621 A CN108198621 A CN 108198621A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- diagnosis
- database
- dicision
- 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.)
- Granted
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The present invention provides a kind of database datas based on neural network to integrate dicision of diagnosis and treatment method, and the method uses nasopharyngeal carcinoma database, and the nasopharyngeal carcinoma database includes clinical large database concept and gene database;Record has the first data formed by clinical alphanumeric information and according to clinical image information extraction and the second data for identifying in the clinic large database concept;The first data, the second data and/or gene data in nasopharyngeal carcinoma database are all with its time attribute;Data of the method in the nasopharyngeal carcinoma database are based on the 5th model under supporting and carry out dicision of diagnosis and treatment;5th model in the first data, the second data and gene data it is a variety of for input, dicision of diagnosis and treatment is provided to the user based on neural network.The present invention can provide the accurately intelligent diagnostics decision service based on artificial intelligence to the user, have wide prospect of the application.
Description
Technical field
The present invention relates to medical fields more particularly to a kind of database data based on neural network to integrate dicision of diagnosis and treatment side
Method.
Background technology
Artificial intelligence is research, exploitation for simulating, extend and extend intelligent theory, method, technology and the application of people
A special kind of skill science of system.Artificial intelligence is a branch of computer science, it attempts to understand the essence of intelligence, and produce
Go out a kind of new intelligence machine that can be made a response in a manner that human intelligence is similar, the research in the field includes robot, language
Speech identification, image identification, natural language processing and expert system etc..Artificial intelligence medical treatment is that its extension in medical field should
With.Artificial intelligence medical treatment at present mainly includes virtual assistant, medical big data, medical image, intelligent sound, body-building biology skill
Art, healthy Lifestyle management, medical treatment search, cancer morning sieve, artificial intelligence chip etc..It is existing the related intelligence of medical treatment occur
System includes Watson robots of IBM Corporation, middle mountain Eye Center " CC-Cruiser congenital cataract artificial intelligence platform "
Deng.
But existing artificial intelligence medical system is just like the intelligence system in this kind of specialized health field of nasopharyngeal carcinoma,
By taking Watson robots as an example, function is relatively complete, can cover a variety of kinds of tumor, but to the current common spy in some China
The cancer training field of color is deeply insufficient.And remove except Watson robots, the system functions such as some domestic functional platforms compared with
To be single, mainly for a certain specific small range disease or inspection item, far from achieving the effect that systematization assisting in diagnosis and treatment.
Nasopharyngeal carcinoma is Guangdong Province's local characteristic cancer kind, and recurrence and transfer are underlying cause of deaths, precisely predicts relapse and metastasis and refers to
It is the key that improve curative effect to lead individualized treatment, and how to carry out based on intelligentized data analysis and thus accurately decision and be
Urgent problem to be solved at present.
Invention content
To solve the above-mentioned problems, the present invention provides a kind of database data synthesis dicision of diagnosis and treatment side based on neural network
Method.
The present invention is realized with following technical solution:
A kind of database data synthesis dicision of diagnosis and treatment method based on neural network, the method use nasopharyngeal carcinoma data
Library, the nasopharyngeal carcinoma database include clinical large database concept and gene database;Being recorded in the clinic large database concept has by facing
The first data that bed alphanumeric information is formed and according to clinical image information extraction and the second data for identifying;Nasopharyngeal carcinoma
The first data, the second data and/or gene data in database are all with its time attribute;
Data of the method in the nasopharyngeal carcinoma database are based on the 5th model under supporting and carry out dicision of diagnosis and treatment;It is described
5th model in the first data, the second data and gene data it is a variety of for input, provided to the user and examined based on neural network
Treat decision.
Further, it further includes and is determined based on the progress diagnosis and treatment of the first model, the second model, third model and/or the 4th model
Plan, first model is using the first data as input, and second model is using the second data as input, and the third model is with base
Because data are input, the 4th model is using data multigroup in nasopharyngeal carcinoma database as input.
Further, the construction method of the 5th model includes:
Extract all or part of feature in different group data;
Higher-dimension is carried out to characteristic feature associated in different groups of data to be abstracted to obtain high dimensional feature;
Complete the filtering to redundancy feature of noise;
All high dimensional features are attached by full articulamentum, so as to generate the 5th model.
Further, it further includes and is adjusted in the 5th model by error backpropagation algorithm combination stochastic gradient descent
Parameter.
Further, it further includes and limits the 5th mould using dropout, leaky relu activation primitives and norm constraint
The adjusting range of parameter in type.
Further, it is finally complete to adjust the 5th model in the prediction efficiency of training set and verification collection by the 5th model
The number and relevant parameter of articulamentum.
Further, the training set includes the first data, the second data and/or gene data.
The beneficial effects of the invention are as follows:
The present invention provides a kind of database data synthesis dicision of diagnosis and treatment method based on neural network, can provide to the user
Accurately intelligent diagnostics decision service based on artificial intelligence, has wide prospect of the application.
Description of the drawings
Fig. 1 is nasopharyngeal carcinoma database schematic diagram provided in an embodiment of the present invention;
Fig. 2 is the forming method flow chart of the first model provided in an embodiment of the present invention;
Fig. 3 is the construction method flow chart of the second model provided in an embodiment of the present invention;
Fig. 4 is the construction method flow chart of third model provided in an embodiment of the present invention;
Fig. 5 is the construction method flow chart of the 4th model provided in an embodiment of the present invention;
Fig. 6 is the method flow diagram provided in an embodiment of the present invention that confluence analysis is carried out to multidimensional data;
Fig. 7 is the construction method flow chart of the 5th model provided in an embodiment of the present invention;
Fig. 8 is cloud system schematic diagram provided in an embodiment of the present invention;
Fig. 9 is analysis decision server schematic diagram provided in an embodiment of the present invention;
Figure 10 is dicision of diagnosis and treatment terminal schematic diagram provided in an embodiment of the present invention;
Figure 11 is application method flow chart in a kind of online assisting in diagnosis and treatment system provided in an embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, the present invention is made into one below in conjunction with attached drawing
Step ground detailed description.
The embodiment of the present invention provides a kind of nasopharyngeal carcinoma artificial intelligence assisting in diagnosis and treatment decision system, to build the dicision of diagnosis and treatment
System, the embodiment of the present invention build nasopharyngeal carcinoma database first.Specifically, in the embodiment of the present invention nasopharyngeal carcinoma database composition
The design of structure is considering based on following research purposes:
Multiple dimensioned isomery association between the macroscopical multi-modality images of further investigation, Clinicopathologic phenotype and microcosmic genotype is closed
System, to establish the interaction relation between clinical phenotypes and microcosmic genotype, so as to complete to nasopharyngeal carcinoma from macroscopic view to microcosmic each
Multiple dimensioned, multi-modal, the big data description of a state, foundation can accurately, full forecast nasopharynx carcinogenesis DISTANT METASTASES IN and answer
The prediction model of hair.
Nasopharyngeal Carcinoma Patients clinical data, image data and gene data are collected, establishes large-scale multi-source heterogeneous multigroup number
It is final to develop nasopharyngeal carcinoma big data high in the clouds diagnosis and treatment analysis system according to library, and precisely treated for auxiliary direction.
In view of this, the nasopharyngeal carcinoma database that is provided in the embodiment of the present invention as shown in Figure 1, including clinical large database concept and
Gene database.
Specifically, record has the first data formed by clinical alphanumeric information and root in the clinical large database concept
According to clinical image information extraction and the second data for identifying.Wherein, the first data can be word and/or digital form, institute
State the second data can be picture and/or visual form, and by number, binary large object, link or it is multimedia in the form of into
Row storage, first data and the second data can analyze the data source of data as a kind of multidimensional, for subsequent diagnosis and treatment
Decision uses.
Specifically, clinical large database concept can include word digital data library, and the word digital data library is used to store
First data, corresponding clinic alphanumeric information include but not limited to check survey report, treatment data and treatment feelings
Condition data.Specifically, the inspection survey report includes medical record data, test rating, pathology, ultrasound, image and/or nuclear medicine
Deng inspection survey report;The treatment data includes but not limited to radiotherapy, chemotherapy, operation, targeting, biological therapy and/or the traditional Chinese medical science
Treatment, the treatment data include but not limited to follow-up prognostic data and therapeutic effect, complication data.
Specifically, clinical large database concept can also include image data base, the second data of described image database purchase,
Corresponding clinical image information includes but not limited to all kinds of clinical images such as image, target of prophylactic radiotherapy, ultrasound, nuclear medicine, pathology.
It is emphasized that the first data, the second data and/or genomics data in nasopharyngeal carcinoma database are all adjoint
Its time attribute records together, and the time attribute is provided with certain effect in various dicision of diagnosis and treatment, during by by each index
Between attribute be included in considering for dicision of diagnosis and treatment, can be provided for survival of patients prognosis according to when the result of decision.
In the case where the data of nasopharyngeal carcinoma database are supported, the dicision of diagnosis and treatment system can be based on a variety of dicision of diagnosis and treatment models pair
Nasopharyngeal carcinoma carries out prediction and decision guidance.
In a feasible embodiment, the dicision of diagnosis and treatment system can include the first diagnosis and treatment module, and described first examines
Nasopharyngeal carcinoma prediction and decision guidance can be provided based on the first model by treating module.First model is with the first data
For data source, the forming method of first model as shown in Fig. 2, including:
S1. the first data are analyzed, the prognosis of factor each in the first data and patient and therapeutic effect is closed
Connection, removal correlation is stronger, the higher factor of synteny.
S2. filter out has the characteristic factor for determining to significantly affect to patient's prognosis, establishes the first model.
S3. verification analysis is carried out to the first model.
First model can be using some or certain several fields in the first data or the first data as input, with patient
Prognostic indicator be output, the prognostic indicator includes but not limited to:N survival rates and tumor recurrence, transfer and/or simultaneously
Send out disease;Prognostic indicator is used to indicate the final final result of patient and the development trend of the nasopharyngeal carcinoma state of an illness.Further, described
Can be obtained in the first data in one model has the index of directive property and the corresponding weight of the index to patient's prognosis.
Further, with the hair of the medical shared data of the continuous expansion and various regions of clinical large database concept data volume
Exhibition, can also be by from itself, either the data of different data center or medical system are introduced into the first model for being obtained in S2
Verification analysis is carried out, and corrects relevant parameter.It, can also school in real time while the content for expanding clinical large database concept continuous
The relevant parameter of positive first model so that the first model constantly improve is to be optimal, so as to for nasopharyngeal carcinoma prediction and certainly
Plan guidance provides more valuable reference.
First model is the model established with the first data bit research object, the first data among clinical position
Acquiring way is relatively simple, has some superiority in data recipient face, therefore, the first model has preferably in clinical position
Application prospect.
In another feasible embodiment, the dicision of diagnosis and treatment system can also include the second diagnosis and treatment module, and described the
Two diagnosis and treatment modules can provide nasopharyngeal carcinoma prediction and decision guidance based on the second model.Second model is with second
Data are data source.The construction method of second model as shown in figure 3, including:
S10. image data is pre-processed.
S20. image is split and three-dimensional reconstruction.
S30. High-throughput quantitative analysis is carried out to image, builds the second model based on image.
Second model can be using some or certain several parts in the second data or the second data as input, with patient
Prognostic indicator be output, the prognostic indicator includes but not limited to:N survival rates and tumor recurrence, transfer and/or simultaneously
Send out disease;Prognostic indicator is used to indicate the final final result of patient and the development trend of the nasopharyngeal carcinoma state of an illness.Further, described
Can be obtained in the second data in two models has the index of directive property and the corresponding weight of the index to patient's prognosis.
Specifically, described image is preferably for nasopharyngeal carcinoma CT/MRI images, image is split and three-dimensional reconstruction after, i.e.,
The High-throughput quantitative analysis of the images such as multi-parameter CT/MRI can be digitized.Concrete analysis content includes but not limited to feature ginseng
The extraction of number data, characteristic parameter data dependence analysis, cluster and statistics Exploration on Characteristics, from large batch of relapse and metastasis nasopharynx
Highly sensitive transfer and relapse image parameters is found in the Multimodal medical image of cancer, with reference to the prognosis feelings such as Patients on Recurrence, transfer
Condition builds the second model based on images such as nasopharyngeal carcinoma CT/MRI.
Further, with the hair of the medical shared data of the continuous expansion and various regions of clinical large database concept data volume
Exhibition, can also be by from itself, either the data of different data center or medical system are introduced into the second mould for being obtained in S30
Type carries out verification analysis, and corrects relevant parameter.Further, while the content of the clinical large database concept of continuous expansion, also
The relevant parameter of the second model can be corrected in real time so that model constantly improve is to be optimal.
Second model is improvement and image technology with medical condition using the second data as the model of research object
It is universal, effect of second data among clinical position becomes increasingly conspicuous, the second data for medical diagnosis directive property also by
Step enhancing, therefore the second model is likewise supplied with preferable application prospect in clinical position.
Further, gene database is for storing gene data, compared to the first data and the second data, gene data
Acquisition difficulty higher, analyze difficulty bigger, but the effect of its prognostic analysis for nasopharyngeal carcinoma is also very prominent.Base
Because the foundation of database can be that following work create conditions:Microcosmic point analyzes the genome sequencing of relapse and metastasis nasopharyngeal carcinoma
Data probe into the genotypic variation in tumorigenesis, apoptotic process, to find potential transfer and relapse driving gene and
Molecular marker, so as to provide important foundation for clinical precisely medical research.
As it can be seen that nasopharyngeal carcinoma database described in the embodiment of the present invention is to incorporate clinical large database concept and gene database
Multi-source heterogeneous multigroup database, the data of storage, which constitute multidimensional, can analyze data.
In another feasible embodiment, the dicision of diagnosis and treatment system can also include third diagnosis and treatment module, and described the
Three diagnosis and treatment modules can provide nasopharyngeal carcinoma prediction and decision guidance based on third model.The third model is with gene
Data are data source.The construction method of the third model as shown in figure 4, including:
S100. genome sequencing is carried out to the nasopharyngeal carcinoma of pairing, transfer stove, recurrence stove.
S200. bioinformatic analysis is carried out to the data after sequencing.
Comprehensive biological bioinformatics analysis is carried out to the data after sequencing, height phase is found in nasopharyngeal carcinoma transfer and relapse sample
Correlation gene target spot, difference molecule and molecular marker find significant transfer, recurrence driving gene and abrupt information and turn
It moves, recurrence related pathways.
S300. the third model based on nasopharyngeal carcinoma gene data is built according to analysis result.
The transfer driving gene and related mutation information obtained to analysis, using internal experiment in vitro, carries out functional verification,
Build the third model based on nasopharyngeal carcinoma gene data.The third model can all or part of gene data be input, with
The prognostic indicator of patient is output, and the prognostic indicator includes but not limited to:N survival rates and tumor recurrence, transfer and/
Or complication;Prognostic indicator is used to indicate the final final result of patient and the development trend of the nasopharyngeal carcinoma state of an illness.Further, in institute
It states to obtain in gene data in third model and there is the index of directive property and the corresponding power of the index to patient's prognosis
Weight.
Further, with the medical shared data of the continuous expansion of gene database data volume and all parts of the country
Development can also will be introduced into the third model progress obtained in S300 from the data at different data center or medical system
Verification analysis, and correct relevant parameter.It further, can also be in real time while the content for constantly expanding gene database
Ground corrects the relevant parameter of third model so that model constantly improve is to be optimal.
The third model is the model established by research object of gene data, with gene technology and correlative study
It is universal, effect of the gene data among clinical position becomes increasingly conspicuous, gene data for medical diagnosis directive property also by
Step enhancing, therefore third model is likewise supplied with preferable application prospect in clinical position.
The data source of first model, the second model and third model is different, and based on a variety of data sources into
Various data sources can be included in the dicision of diagnosis and treatment of nasopharyngeal carcinoma by row to be considered, and is referred to so as to provide more comprehensive diagnosis and treatment for patient
It leads.Based on this cognition, in another feasible embodiment, the dicision of diagnosis and treatment system can also include the 4th diagnosis and treatment mould
Block, the 4th diagnosis and treatment module can provide nasopharyngeal carcinoma prediction and decision guidance based on the 4th model.4th mould
Type can be using a variety of data in the first data, the second data and third data in nasopharyngeal carcinoma database as research object
User carries out dicision of diagnosis and treatment guidance.The construction method of wherein the 4th model as shown in figure 5, including:
P1. confluence analysis is carried out to multigroup data, establishes the Data Storage Models after quality evaluation.
Specifically, multigroup data of learning can be a variety of in the first data, the second data and gene data, wherein the
One data, the second data and gene data are all a group data.
P2. comprehensive system probes into the relevance between multigroup data, establishes based on multigroup the 4th mould for learning Data Integration
Type.
The 4th model includes public discovery model in the embodiment of the present invention, can specifically be based on sparse expression the Theory Construction
The public discovery model.
It, can be in tensor expression and the frame of expression matrix in P2 implementation procedures in another feasible embodiment
Under frame, the characteristic model described towards tensor is proposed;Establish the tensor pairing energy equation based on Prior Knowledge Constraints;Analysis is special
Sign tensor describes method, Prior Knowledge Constraints design, the internal relation of tensor matching method and local Common Substructure;For height
The data of noise pollution, the robustness of test model;For mass data, establish efficient restricted problem and decompose mechanism, to
Rapid Optimum solves, so as to finally obtain public discovery model.
Specifically, the embodiment of the present invention proposes the method for carrying out confluence analysis to multigroup data, as shown in fig. 6, packet
It includes:
First, it establishes towards multigroup matrix decomposition model for learning data, multidimensional data is decomposed in same base space,
Interactive relationship between coefficient of analysis matrix, so as to find the incidence relation between multigroup data.
Specifically, correlation rule between research genotype and phenotype is established, important gene target spot and image phenotype is found, swells
Dynamically associating between the factors such as knurl stage and step.
Secondly, it establishes towards multigroup tensor resolution model for learning data, by using the side in three implicit variable base spaces
Method decomposes multigroup data in three base spaces, extracts independently of multigroup centronucleus tensor for learning data, research core
Public relation between amount, to find its public association.
Specifically, it studies under different constraints, the resolution characteristic in implicit variable base space is designed towards multiple constraint centronucleus
The correlation function of tensor.
Again, it establishes the tensor described based on tensor and matches model, the tensor property for inquiring into single group data describes method,
The influence that research tensor property dimension and quantitative description match tensor;The matching method of design feature tensor, structure is based on first
Test the tensor pairing model of knowledge constraints.
Finally, it establishes public module for above-mentioned analysis result and finds model.
Based on the research method of above-mentioned public discovery model, an embodiment of the present invention provides originality Integrated Models to optimize letter
Number, the Integrated Models majorized function are the core content of public discovery model, for being picked out in multigroup data to nose
Pharynx cancer patient's prognosis has the index of directive property:
Subjectto:Xi,j∈{0,1}
XI1≤b1
XTI2≤b2
Wherein H expressions group data S1And S2Between similarity, X represent allocation matrix,Represent data S1It is internal special
Levy incidence relation,Represent data S2Internal feature incidence relation, it is similar that formula first item represents that allocation matrix X meets as possible
The distribution of matrix H is spent, Section 2 and Section 3 represent to have incorporated data S respectively1With data S2Priori, improve accuracy rate,
It is openness that Section 4 represents that allocation matrix has so that can preferably explain data S1With data S2Between incidence relation.
Wherein LMTo normalize Laplacian Matrix, it is expressed as below
According to Laplacian Matrix property, the following formula can be obtained
Above-mentioned optimization function is solved eventually by gradient descent method, obtains Data Integration as a result, shown in following algorithm 1
Algorithm 1
Input:M1,M2,WM1,WM2
1:Initialize λ1,λ2,β,X(0),γ(1),M,ρ>1,δ∈(0,1),ε>0, k=1
2:Normalize M1,M2
3:The following formula update allocation matrix X is solved using gradient descent method(k)
4:M=ρ M are set, if
5:Update γ(k+1)=min (γ(k)-Mg(X(k)),0)
6:K=k+1 is set
7:Repeat step 3,4,5,6 until | | g (X(k))||<ε
Output:X
Further, have on the basis of directive property index based on the public discovery model acquisition prognosis, in order to
Comprehensive treatment effect is promoted, also is able to weight of the various directive property indexs in Index for diagnosis being included in the 4th model and examine
Amount.The method for obtaining weight based on the 4th model has very much, and the embodiment of the present invention generally lifts three feasible embodiments.
In a feasible embodiment, the output of the 4th model can be simply represented as P=K1A1+K2A2+
K3A3+……+KmAm.P is in A1+A2+A3+ ... under the conditions of+Am, without all kinds of existence such as progression of disease in 5-year Survival/5 year
Possible probability of happening.And the value of the data acquisition A1 ... Am in nasopharyngeal carcinoma database, you can obtain various directive property
Weight of the index in Index for diagnosis.
In another feasible embodiment, the output P of the 4th model can be expressed as to multiple index comprehensive effects
As a result, and the comprehensive function including each prognosis have directive property index act on simultaneously, stage by stage act on and mix make
With.By analyzing the data acquisition in nasopharyngeal carcinoma database to various directive property indexs in Index for diagnosis in this representation method
In weight.
It, can the research based on the first model, the second model and/or third model in another feasible embodiment
As a result, obtain weight of the various directive property indexs in Index for diagnosis.In this embodiment, the structure of the 4th model includes
Following step:
The index for having directive property for prognosis is picked out according to public discovery model;
Judge that the index belongs to the first data, the second data or gene data;
If the index belongs to the first data, the weight of the index is obtained according to the first model;
If the index belongs to the second data, the weight of the index is obtained according to the second model;
If the index belongs to gene data, the weight of the index is obtained according to third model;
After the weight of the index to be obtained for all referring to tropism, dicision of diagnosis and treatment is carried out according to the weighted results of index and its weight
Guidance.
4th model realizes that the dicision of diagnosis and treatment of generalization creates condition for dicision of diagnosis and treatment system, it is clear that in dicision of diagnosis and treatment
System in actual use, needs constantly to obtain the feedback of doctor and patient, with reference to the data and document of continuous renewal,
It constantly corrects public module and finds model, the first model, the second model and third model, so as to be artificial intelligence assisting in diagnosis and treatment
More power-assisteds are provided.
More comprehensively medical treatment result is obtained in order to be based on nasopharyngeal carcinoma database, in another feasible embodiment, institute
The 5th diagnosis and treatment module can also be included by stating dicision of diagnosis and treatment system, and the 5th diagnosis and treatment module can provide nose based on the 5th model
Pharynx cancer prediction and decision guidance.5th model equally can be using data multigroup in nasopharyngeal carcinoma database as data
Source, different from the construction method of the 4th model, the 5th model provides dicision of diagnosis and treatment to the user based on neural network.Wherein the 5th mould
The construction method of type as shown in fig. 7, comprises:
P10 extracts different groups and learns in data for the valuable feature of generation of the 5th model.
Extracting method specifically can there are many, can also refer to the 4th model in public discovery model carry out feature
Extraction.
P20 carries out higher-dimension to characteristic feature associated in different groups of data and is abstracted to obtain high dimensional feature.
P30 completes the filtering to redundancy feature of noise.
In above three step, bottom can be extracted to height using convolutional neural networks to the first data and gene data
The feature representation of layer;Useful feature is excavated using integrated approach to the second data, and passes through traditional logistic regression and combines not
Same norm constraint filters unrelated interference characteristic.
It is abstracted to obtain high dimensional feature, and completion pair by carrying out characteristic feature associated in different groups of data higher-dimension
The filtering of redundancy feature of noise;So as to create conditions for P40.
All high dimensional features are attached, so as to generate the 5th model by P40 by full articulamentum.
Specifically, all high dimensional features can be attached by full articulamentum, generate dicision of diagnosis and treatment model, when with
When the data volume newly collected of training pattern reaches regulation magnitude, model training, Optimized model parameter so that mould are restarted
The continuous self-teaching of type, self adjustment, self-perfection, so as to form the 5th model.
Specifically, the 5th model of the embodiment of the present invention can be by establishing convolutional neural networks and convolutional neural networks
It practises and realizing.
In the building process of the 5th model, the data in nasopharyngeal carcinoma database are collected by a full articulamentum
Into by error backpropagation algorithm combination stochastic gradient descent come adjusting parameter, and using dropout, leaky relu swash
Function living and norm constraint limit the adjusting range of parameter, ensure the openness of feature, avoid the situation of over-fitting, so as to
It is further ensured that the efficiency of the 5th model.Model is adjusted most in the prediction efficiency of training set and verification collection by the 5th model
The number and relevant parameter of full articulamentum afterwards, to provide valuable diagnostic comments for doctor.
4th model and the 5th model are with multigroup model for learning data bit research object, can use nasopharyngeal carcinoma number comprehensively
The more comprehensive result of decision is desirably to obtain according to the content in library, the first model, the second model and third model are with single group number
According to the model for research object, diagnosis and treatment can be carried out on the basis of a certain kind organizes data, in a kind of preferred embodiment
In, the first model, the second model, third model, the 4th model and the 5th model can have or preferentially exist one kind parallel
Or it is a variety of, so as to provide more good dicision of diagnosis and treatment service to the user from multiple angles.
On the basis of the above, the embodiment of the present invention further provides for a kind of nasopharyngeal carcinoma artificial intelligence auxiliary and examines
The logical architecture of decision system is treated, nasopharyngeal carcinoma artificial intelligence assisting in diagnosis and treatment decision system is specifically as follows a kind of online diagnosis and treatment system
System, specifically, the online diagnosis and therapy system can be a kind of cloud system, the data supporter as the online diagnosis and therapy system
(nasopharyngeal carcinoma database) can also be laid beyond the clouds, it is clear that be more advantageous to Data Integration using cloud storage and diagnosis and treatment are analyzed.
The dicision of diagnosis and treatment system as shown in figure 8, can include setting dicision of diagnosis and treatment server beyond the clouds and with it is described
Dicision of diagnosis and treatment server communication connection dicision of diagnosis and treatment terminal, the dicision of diagnosis and treatment system specifically can use B-S frameworks or
C-S frameworks, the dicision of diagnosis and treatment terminal are sent out to the dicision of diagnosis and treatment server cluster corresponding to described in response to user instruction
The request of data of user instruction;The dicision of diagnosis and treatment server in response to the request of data, generation data respond and incite somebody to action described in
Data response is transmitted to the dicision of diagnosis and treatment terminal, in order to which the dicision of diagnosis and treatment terminal corresponds to the user to user's displaying
The result of instruction.
Further, in order to provide consulting services to the user comprehensively, the dicision of diagnosis and treatment server is specifically as follows service
Device cluster (dicision of diagnosis and treatment server cluster), the dicision of diagnosis and treatment server cluster include user's interactive server, analysis decision
Server and data server, user's interactive server are used to be used in combination with the progress data interaction of dicision of diagnosis and treatment terminal
Family manages, and for providing dicision of diagnosis and treatment service to the user, the data server is used for into line number the analysis decision server
According to processing, it is provided with nasopharyngeal carcinoma database and interacts.The analysis decision server is as shown in figure 9, preferably include
One diagnosis and treatment module, the second diagnosis and treatment module, third diagnosis and treatment module, the 4th diagnosis and treatment module and the 5th diagnosis and treatment module, first diagnosis and treatment
Module, the second diagnosis and treatment module, third diagnosis and treatment module, the function of the 4th diagnosis and treatment module and the 5th diagnosis and treatment module are as previously mentioned, herein
No longer superfluous words.
Further, in the dicision of diagnosis and treatment terminal, user's registration is provided with, data management, inquiry, annotates, compare, examining
The functions such as rope, displaying.Specifically, the dicision of diagnosis and treatment terminal includes as shown in Figure 10:
User registration module, for being registered to the dicision of diagnosis and treatment server cluster, in order to which user uses diagnosis and treatment
The various services that policy server cluster provides.
Data management module, for managing the personal data of user.
Enquiry module inquires target data for user to dicision of diagnosis and treatment server cluster.
Annotations module is marked for user in display interface.
Comparison module, for being compared, and identify comparison result in order to which user quickly has found for similar data
Difference between similar data.
Module is retrieved, for user's fast search target data.
Display module, for carrying out data visualization.
The dicision of diagnosis and treatment terminal can be to be mounted with the arbitrary smart machine of the software interacted with dicision of diagnosis and treatment server,
Such as the equipment such as PC, mobile phone or tablet.
In the dicision of diagnosis and treatment terminal, changing interface module is further included, the changing interface module is used to step on according to user
The difference of record identity provides different display interfaces to the user.Specifically, according to the identity difference for using user, changing interface mould
Block is also additionally provided different functions.If user identity is patient user, provides a user and access data, check diagnostic result
With the functions such as therapeutic scheme;If user identity is clinician user, user, which provides, to be accessed data, makes diagnosis decision and prognosis
The functions such as assessment.
It specifically, can be by, using easy friendly operation interface, high in the clouds being divided in Website front-end in B-S frameworks
The medical treatment result visualization of analysis.The visualization can specifically realize by the visualization model of dicision of diagnosis and treatment terminal, it is described can
Include data early warning unit, chart linkage unit and comparison split cells depending on changing module.Clinical speciality doctor is not only facilitated to operate,
And clear medical image display interface and case-data display interface are additionally provided, it can be with to clinical speciality doctor
Preferably make assessment and diagnosis.
The online diagnosis and therapy system is not only that clinical speciality doctor provides recurrent nasopharyngeal carcinoma transfer stage and step tentative prediction
As a result, it additionally provides management, annotate, compare, retrieve and show more than PB magnitude group data and clinical information function.It is described
Line diagnosis and therapy system can provide direction guidance for gene target, image tumor region, accurate Personalized medicine, be doctor and patient
Medical diagnosis on disease is provided and visualizes platform, a reliable and effective model system is provided for tumor research treatment, is also other
The source that tumor research offer can refer to.
Specifically, on the basis of the above, the embodiment of the present invention provides a kind of online diagnosis and therapy system application method and makees
For diagnosis and treatment example, as shown in figure 11, including:
S101. clinical alphanumeric information, clinical image information and/or gene data are obtained, and combines the time of data
Attribute is included in nasopharyngeal carcinoma database.
Specifically, database table structure can be as illustrated in chart 1:
Table 1
S102. selection can be used in the dicision of diagnosis and treatment model of decision, and carry out survival region, complication, analysis using it.
Specifically, the dicision of diagnosis and treatment model for decision can be the first model, the second model, third model, the
It is one or more in four models and the 5th model.
In this step, the model for decision can be the 4th model, and the output of the 4th model has finger to be a kind of to prognosis
The weighted results of the index of tropism temporarily can simply be represented as P=K1A1+K2A2+K3A3+ ...+KmAm.P is in A1+A2+
Under the conditions of A3+ ...+Am, without the possible probability of happening of all kinds of existence such as progression of disease in 5-year Survival/5 year.P is bigger, it was demonstrated that suffers from
It is obtained under the conditions of person is in A1+A2+A3+ ...+Am more long.Alignment graph model etc. can be established to be used to predict A1+A2+A3+ ...+Am
The Survival datas such as the probability survived how many years.
And wherein, according to the size of disparity items Ai COEFFICIENT Ks i, directionality adjustment can be carried out to Ai, to reach more preferably raw
Deposit benefit.
S103. according to having model, with reference to current case, dicision of diagnosis and treatment reference scheme is obtained, and obtain follow up data.
S104. with the addition of new case, constantly learn and optimize existing dicision of diagnosis and treatment model.
In the specific implementation process of the embodiment of the present invention, following achievement can be obtained:
(1) in terms of patient:
Patient fills in itself essential information after logging in, input coherence check, inspection result can obtain relevant disease and examine
The related contents such as points for attention during disconnected, prognosis estimation, treatment.And it with therapeutic advance, is inputted, obtained different according to its difference
It is recommended that.
(2) in terms of researcher
According to data with existing library, the correlation models such as treatment recommendations, prognosis are established.And information is inputted according to patient, it is formed new
Big data, and Parameters in Mathematical Model is corrected, obtains the model that can most react truth.
Data information of both clinical data and gene data is integrated, to find macrophenotypic --- microcosmic genotype
The marker of multilevel lower highlights correlations so as to establish contact of the gene target to image phenotype, and then is established based on big data
The accurate medical system of artificial intelligence.
In addition, following implementation steps in the embodiment of the present invention are put forward for the first time in the related art:
(1) structure has multigroup database of time attribute, establish it is multigroup learn according to when prognosis and complication etc. predict mould
Type.
(2) highly sensitive transfer and relapse image mark is found from the Multimodal medical image of batch relapse and metastasis nasopharyngeal carcinoma
Will;
(3) highly relevant gene target is found in nasopharyngeal carcinoma transfer and relapse sample using genomics bioassay technique
Point, difference molecule and molecular marker;
(4) the multilevel isomeric datas such as clinical alphanumeric information, clinical image information and gene data are integrated, to send out
The marker of the multilevel lower highlights correlations of existing macrophenotypic-sight genotype, so as to establish gene target to the pass of image phenotype
Connection;
(5) build up the database of the extensive different levels group data of nasopharyngeal carcinoma, have developed collection manage, annotate, comparing,
Retrieval, presentation group data and relapse and metastasis are by stages and prediction is in advance after the online diagnosis and therapy system of one.
The above disclosure is only the preferred embodiments of the present invention, cannot limit the right model of the present invention with this certainly
It encloses, therefore equivalent variations made according to the claims of the present invention, is still within the scope of the present invention.
Claims (7)
- A kind of 1. database data synthesis dicision of diagnosis and treatment method based on neural network, which is characterized in that the method uses nose Pharynx cancer database, the nasopharyngeal carcinoma database include clinical large database concept and gene database;Remember in the clinic large database concept Record have the first data formed by clinical alphanumeric information and according to clinical image information extraction and identify second number According to;The first data, the second data and/or gene data in nasopharyngeal carcinoma database are all with its time attribute;Data of the method in the nasopharyngeal carcinoma database are based on the 5th model under supporting and carry out dicision of diagnosis and treatment;Described 5th Model in the first data, the second data and gene data it is a variety of for input, diagnosis and treatment are provided to the user based on neural network and are determined Plan.
- 2. dicision of diagnosis and treatment method is integrated according to a kind of database data based on neural network described in claim 1, it is special Sign is, further includes and carries out dicision of diagnosis and treatment based on the first model, the second model, third model and/or the 4th model, described first Model is using the first data as input, and second model is using the second data to input, and the third model is using gene data to be defeated Enter, the 4th model is using data multigroup in nasopharyngeal carcinoma database as input.
- 3. a kind of database data synthesis dicision of diagnosis and treatment method based on neural network according to claim 2, feature It is:The construction method of 5th model includes:Bottom is extracted to high-rise feature representation using convolutional neural networks to the first data and gene data;Feature mining is carried out using integrated approach to the second data, and it is special to pass through the filtering interference of logistic regression combination norm constraint Sign;Higher-dimension is carried out to characteristic feature associated in different groups of data to be abstracted to obtain high dimensional feature;Obtained high dimensional feature is attached by full articulamentum, generates dicision of diagnosis and treatment model;When training set quantity is to preset value, the training of dicision of diagnosis and treatment model is restarted, so that model passes through self-teaching Optimize relevant parameter, so as to form the 5th model.
- 4. a kind of database data synthesis dicision of diagnosis and treatment method based on neural network according to claim 3, feature It is:It further includes and the parameter in the 5th model is adjusted by error backpropagation algorithm combination stochastic gradient descent.
- 5. a kind of database data synthesis dicision of diagnosis and treatment method based on neural network according to claim 3, feature It is:It further includes and limits the parameter in the 5th model using dropout, leaky relu activation primitives and norm constraint Adjusting range.
- 6. a kind of database data synthesis dicision of diagnosis and treatment method based on neural network according to claim 5, feature It is:The number of the last full articulamentum of 5th model is adjusted in the prediction efficiency of training set and verification collection by the 5th model And relevant parameter.
- 7. a kind of database data synthesis dicision of diagnosis and treatment method based on neural network according to claim 6, feature It is:The training set includes the first data, the second data and/or gene data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810047166.7A CN108198621B (en) | 2018-01-18 | 2018-01-18 | Database data comprehensive diagnosis and treatment decision method based on neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810047166.7A CN108198621B (en) | 2018-01-18 | 2018-01-18 | Database data comprehensive diagnosis and treatment decision method based on neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108198621A true CN108198621A (en) | 2018-06-22 |
CN108198621B CN108198621B (en) | 2022-03-08 |
Family
ID=62590125
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810047166.7A Active CN108198621B (en) | 2018-01-18 | 2018-01-18 | Database data comprehensive diagnosis and treatment decision method based on neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108198621B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109087702A (en) * | 2018-08-03 | 2018-12-25 | 厦门大学 | Four methods of diagnosis characterization information fusion method for the analysis of Chinese medicine health status |
CN109102494A (en) * | 2018-07-04 | 2018-12-28 | 中山大学中山眼科中心 | A kind of After Cataract image analysis method and device |
CN109887589A (en) * | 2019-02-21 | 2019-06-14 | 吉林禾熙科技开发有限公司 | Artificial intelligence medical treatment big data system |
CN109903846A (en) * | 2019-02-21 | 2019-06-18 | 吉林禾熙科技开发有限公司 | Screening lung cancer system and method based on medical big data |
CN110277165A (en) * | 2019-06-27 | 2019-09-24 | 清华大学 | Aided diagnosis method, device, equipment and storage medium based on figure neural network |
CN110791564A (en) * | 2018-10-10 | 2020-02-14 | 杭州翱锐基因科技有限公司 | Method and apparatus for analyzing early cancer |
CN111128328A (en) * | 2019-10-25 | 2020-05-08 | 中山大学肿瘤防治中心(中山大学附属肿瘤医院、中山大学肿瘤研究所) | Nasopharyngeal carcinoma structured image report and data processing system and method |
CN111508618A (en) * | 2020-04-13 | 2020-08-07 | 南京巨鲨显示科技有限公司 | Perioperative period artificial intelligence auxiliary platform system |
WO2020205140A1 (en) | 2019-04-02 | 2020-10-08 | Kpn Innovations, Llc | Methods and systems for utilizing diagnostics for informed vibrant constitutional guidance |
CN111768818A (en) * | 2020-06-05 | 2020-10-13 | 上海市浦东医院(复旦大学附属浦东医院) | Artificial intelligence-based pelvis fracture first-aid decision-making system and method |
CN112017785A (en) * | 2020-11-02 | 2020-12-01 | 平安科技(深圳)有限公司 | Disease risk prediction system, method, device, equipment and medium |
CN112270660A (en) * | 2020-09-30 | 2021-01-26 | 四川大学 | Nasopharyngeal carcinoma radiotherapy target area automatic segmentation method based on deep neural network |
CN113207307A (en) * | 2019-12-03 | 2021-08-03 | 克里克疗法有限公司 | Apparatus, system, and method for determining demographic information to facilitate mobile application user engagement |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102827941A (en) * | 2012-09-06 | 2012-12-19 | 温州医学院附属第一医院 | Gene chip testing kit for recurrent nasopharyngeal carcinoma |
CN107247881A (en) * | 2017-06-20 | 2017-10-13 | 北京大数医达科技有限公司 | A kind of multi-modal intelligent analysis method and system |
CN107273685A (en) * | 2017-06-14 | 2017-10-20 | 广东工业大学 | A kind of data analysing method of multi-modal big data for clinical disease |
CN107463774A (en) * | 2017-07-21 | 2017-12-12 | 温馨港网络信息科技(苏州)有限公司 | Analysis on the health status Forecasting Methodology and system based on big data |
-
2018
- 2018-01-18 CN CN201810047166.7A patent/CN108198621B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102827941A (en) * | 2012-09-06 | 2012-12-19 | 温州医学院附属第一医院 | Gene chip testing kit for recurrent nasopharyngeal carcinoma |
CN107273685A (en) * | 2017-06-14 | 2017-10-20 | 广东工业大学 | A kind of data analysing method of multi-modal big data for clinical disease |
CN107247881A (en) * | 2017-06-20 | 2017-10-13 | 北京大数医达科技有限公司 | A kind of multi-modal intelligent analysis method and system |
CN107463774A (en) * | 2017-07-21 | 2017-12-12 | 温馨港网络信息科技(苏州)有限公司 | Analysis on the health status Forecasting Methodology and system based on big data |
Non-Patent Citations (1)
Title |
---|
华贻军,洪明晃等: "应用人工神经网络方法预测鼻咽癌患者5年生存状态", 《肿瘤学杂志》 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109102494A (en) * | 2018-07-04 | 2018-12-28 | 中山大学中山眼科中心 | A kind of After Cataract image analysis method and device |
CN109087702B (en) * | 2018-08-03 | 2021-07-16 | 厦门大学 | Four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis |
CN109087702A (en) * | 2018-08-03 | 2018-12-25 | 厦门大学 | Four methods of diagnosis characterization information fusion method for the analysis of Chinese medicine health status |
CN110791564A (en) * | 2018-10-10 | 2020-02-14 | 杭州翱锐基因科技有限公司 | Method and apparatus for analyzing early cancer |
CN110791564B (en) * | 2018-10-10 | 2022-07-08 | 杭州翱锐基因科技有限公司 | Method and apparatus for analyzing early cancer |
CN109887589A (en) * | 2019-02-21 | 2019-06-14 | 吉林禾熙科技开发有限公司 | Artificial intelligence medical treatment big data system |
CN109903846A (en) * | 2019-02-21 | 2019-06-18 | 吉林禾熙科技开发有限公司 | Screening lung cancer system and method based on medical big data |
EP3948760A4 (en) * | 2019-04-02 | 2022-12-28 | KPN Innovations, LLC | Methods and systems for utilizing diagnostics for informed vibrant constitutional guidance |
WO2020205140A1 (en) | 2019-04-02 | 2020-10-08 | Kpn Innovations, Llc | Methods and systems for utilizing diagnostics for informed vibrant constitutional guidance |
CN110277165A (en) * | 2019-06-27 | 2019-09-24 | 清华大学 | Aided diagnosis method, device, equipment and storage medium based on figure neural network |
CN110277165B (en) * | 2019-06-27 | 2021-06-04 | 清华大学 | Auxiliary diagnosis method, device, equipment and storage medium based on graph neural network |
CN111128328A (en) * | 2019-10-25 | 2020-05-08 | 中山大学肿瘤防治中心(中山大学附属肿瘤医院、中山大学肿瘤研究所) | Nasopharyngeal carcinoma structured image report and data processing system and method |
CN113207307A (en) * | 2019-12-03 | 2021-08-03 | 克里克疗法有限公司 | Apparatus, system, and method for determining demographic information to facilitate mobile application user engagement |
CN111508618A (en) * | 2020-04-13 | 2020-08-07 | 南京巨鲨显示科技有限公司 | Perioperative period artificial intelligence auxiliary platform system |
CN111508618B (en) * | 2020-04-13 | 2023-09-15 | 南京巨鲨显示科技有限公司 | Manual intelligent auxiliary platform system for perioperative period |
CN111768818A (en) * | 2020-06-05 | 2020-10-13 | 上海市浦东医院(复旦大学附属浦东医院) | Artificial intelligence-based pelvis fracture first-aid decision-making system and method |
CN112270660A (en) * | 2020-09-30 | 2021-01-26 | 四川大学 | Nasopharyngeal carcinoma radiotherapy target area automatic segmentation method based on deep neural network |
CN112017785A (en) * | 2020-11-02 | 2020-12-01 | 平安科技(深圳)有限公司 | Disease risk prediction system, method, device, equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN108198621B (en) | 2022-03-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108198621A (en) | A kind of database data synthesis dicision of diagnosis and treatment method based on neural network | |
CN108320807A (en) | A kind of nasopharyngeal carcinoma artificial intelligence assisting in diagnosis and treatment decision cloud system | |
CN108206056A (en) | A kind of nasopharyngeal carcinoma artificial intelligence assisting in diagnosis and treatment decision terminal | |
CN108335756A (en) | The synthesis dicision of diagnosis and treatment method in nasopharyngeal carcinoma database and based on the data library | |
Tang et al. | CapSurv: Capsule network for survival analysis with whole slide pathological images | |
Ghazal et al. | Feature optimization and identification of ovarian cancer using internet of medical things | |
CN109830303A (en) | Clinical data mining analysis and aid decision-making method based on internet integration medical platform | |
Huang et al. | Integration of patch features through self-supervised learning and transformer for survival analysis on whole slide images | |
US7809660B2 (en) | System and method to optimize control cohorts using clustering algorithms | |
Puppala et al. | METEOR: an enterprise health informatics environment to support evidence-based medicine | |
CN109599157B (en) | Accurate intelligent diagnosis and treatment big data system | |
CN108520778A (en) | A kind of screening lung cancer system and method based on big data | |
CN109785927A (en) | Clinical document structuring processing method based on internet integration medical platform | |
CN114003734B (en) | Knowledge system and knowledge map system of breast cancer risk factors and construction method | |
Cui et al. | L 2, 1-GRMF: an improved graph regularized matrix factorization method to predict drug-target interactions | |
Liu et al. | Big medical data decision-making intelligent system exploiting fuzzy inference logic for prostate cancer in developing countries | |
WO2021143781A1 (en) | Multi-center synergetic cancer prognosis prediction system based on multi-source migration learning | |
CN108320797A (en) | A kind of nasopharyngeal carcinoma database and based on the data the synthesis dicision of diagnosis and treatment method in library | |
Srivastava et al. | Complex predictive analysis for health care: a comprehensive review | |
CN108335748A (en) | A kind of nasopharyngeal carcinoma artificial intelligence assisting in diagnosis and treatment policy server cluster | |
Sethi et al. | Long Short-Term Memory-Deep Belief Network based Gene Expression Data Analysis for Prostate Cancer Detection and Classification | |
Dai et al. | Patient similarity: methods and applications | |
CN117457065A (en) | Method and system for identifying phenotype-associated cell types based on single-cell multi-set chemical data | |
CN117038096A (en) | Chronic disease prediction method based on low-resource medical data and knowledge mining | |
Lu et al. | M 2 F: A Multi-modal and Multi-task Fusion Network for Glioma Diagnosis and Prognosis |
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 |