CN107863147A - The method of medical diagnosis based on depth convolutional neural networks - Google Patents
The method of medical diagnosis based on depth convolutional neural networks Download PDFInfo
- Publication number
- CN107863147A CN107863147A CN201711002335.7A CN201711002335A CN107863147A CN 107863147 A CN107863147 A CN 107863147A CN 201711002335 A CN201711002335 A CN 201711002335A CN 107863147 A CN107863147 A CN 107863147A
- Authority
- CN
- China
- Prior art keywords
- diagnosed
- electronic health
- health record
- medical
- term vector
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses the method for the medical diagnosis based on depth convolutional neural networks, this method includes：Obtain term vector matrix corresponding to electronic health record to be diagnosed；By term vector Input matrix corresponding to electronic health record to be diagnosed into the depth convolutional neural networks model built in advance, the characteristic vector of electronic health record to be diagnosed is obtained；The characteristic vector that the electronic health record of diagnosis is treated using grader is classified, and obtains the P of each illness corresponding to electronic health record to be diagnosed.Convolutional neural networks are applied to medical electronics case history text semantic and understand and carry out assisted medical diagnosis by this method, can effectively overcome defect possessed by the method for rule-based extraction and matching.
Description
Technical field
The present invention relates to medical information technical field, more particularly to medical diagnosis based on depth convolutional neural networks
Method.
Background technology
Clinical decision support (Clinical decision support system, CDSS) refers to related, system
Clinical knowledge and patient basis and state of an illness information, strengthen medical treatment related decision-making and action, improve quality of medical care and doctor
Treat service level.CDSS is the important means for lifting quality of medical care, and its basic goal is to assess and improve quality of medical care, subtract
Few malpractice, control the expenditure of medical expense.
At present, most CDSS are all based on Rule Extraction and matching to realize in the world, using following technical side
Case：Based on clinical database, by being collected to information, arranging, classifying, filtering, processing etc. and to set up logic association knowledge
Point；Using warning prompting, in groups information button, doctor's advice (doctor's advice set meal), document management and related data expression form；To disease
Disease is diagnosed, treated, being nursed, being performed the operation, the decision support of the rational use of medicines etc.；There is provided and build for clinician's diagnoses and treatment
View, remind, alarm, calculating, the decision support in terms of prediction；
It is rule-based extraction and matching method have its it is intrinsic the defects of, specifically have it is following some：
1st, the problem of semantic ambiguity be present.Such as description headache when may have a variety of different words such as " headache " and " headache " but
It is with semantic description method, needs to include all possible description as much as possible when building knowledge base, this can cause knowledge
The redundancy and efficiency in storehouse are low, while as comprising all possible situation matching accuracy rate could not be caused to decline.
2nd, the case species that each hospital department is run into is various and difference is larger, it is necessary to hundreds and thousands of to so more section office
Individual disease builds knowledge base, and exception is cumbersome, is unfavorable for managing and safeguards, causes efficiency extremely low；
3rd, the structure in knowledge based storehouse carries out assisted medical diagnosis, has the suspicion of " empty talk " unavoidably, it should be understood that medical diagnosis
Clinical practice and experience are extremely relied on, with description structure rule on book and then clinical practice is instructed, can play and put the cart before the horse
Effect, it is difficult to reach the purpose of auxiliary diagnosis.
Therefore, a kind of method of the medical diagnosis based on depth convolutional neural networks turns into technical problem urgently to be resolved hurrily.
The content of the invention
The purpose of the present invention is intended at least solve one of above-mentioned technical problem to a certain extent.
Therefore, first purpose of the present invention is the side of the medical diagnosis based on depth convolutional neural networks proposed
Convolutional neural networks are applied to medical electronics case history text semantic and understand and carry out assisted medical diagnosis to have by method, this method
Effect overcomes defect possessed by the method for rule-based extraction and matching.This method can effectively solve the problems, such as " semantic gap ",
Effectively eliminate influence caused by synonymous different words etc. during electronic health record is write；This method can be directed to each section office's disease it is various and
The larger situation of difference, it is not necessary to build corresponding rule and matching algorithm for each disease, unified model frame can be built
Frame, then with the historical data training pattern of each disease, it is possible to as long as multiple diseases can just be examined in advance by reaching a model
Effect, it is especially suitable for managing and safeguards, scalability is also very strong；This method does not need artificial design rule and feature, model
All features and rule learnt come from clinical a large amount of historical datas, are entirely to instruct clinic to determine with clinical history data
Plan, there is very strong practical guided significance compared to rule-based aided diagnosis method.
To achieve these goals, the medical diagnosis based on depth convolutional neural networks of first aspect present invention embodiment
Method, including：
Obtain term vector matrix corresponding to electronic health record to be diagnosed；
By term vector Input matrix corresponding to the electronic health record to be diagnosed to the depth convolutional Neural net built in advance
In network model, characteristic vector corresponding to electronic health record to be diagnosed is obtained；
The characteristic vector of the electronic health record to be diagnosed is classified using grader, electricity to be diagnosed described in acquisition
The P of each illness corresponding to sub- case history.
Method as described above, it is described to obtain term vector matrix corresponding to electronic health record to be diagnosed, including：
The electronic health record for treating diagnosis carries out at least one of information filtering, screening, participle, statistics operation, and acquisition treats
Diagnose each medical vocabulary of case history；
Term vector corresponding to the medical vocabulary of case history to be diagnosed described in being obtained in the default term vector database, its
In, the corresponding relation of medical vocabulary and term vector is preserved in the default term vector database；
It is right according to term vector generation follow-up power-off son disease corresponding to the medical vocabulary of each electronic health record to be diagnosed
The term vector matrix answered.
Method as described above, it is described obtain term vector matrix corresponding to electronic health record to be diagnosed before, including：
Obtain each medical vocabulary in medical dictionary；
Medical vocabulary in the medical dictionary is input in the Word2Vec models pre-established, obtains the medical treatment
Term vector corresponding to vocabulary；
Term vector corresponding with the medical vocabulary is formed into term vector sample, by the term vector Sample preservation default
In term vector database.
Method as described above, before each medical vocabulary obtained in medical dictionary, including：
Obtain multiple electronic health records diagnosed；
Information filtering is carried out to each electronic health record diagnosed using Information Filtering Technology, obtains medical word finder
Close；
The word frequency of each medical vocabulary in the medical lexical set is counted, each medical treatment is screened according to setting screening rule
Vocabulary, the medical dictionary is established according to the selection result.
Method as described above, it is described obtain term vector matrix corresponding to electronic health record to be diagnosed before, including：
Term vector matrix corresponding to multiple electronic health records diagnosed is obtained, by corresponding to the electronic health record diagnosed
Term vector matrix is as training sample；
The training sample is trained, builds the depth convolutional neural networks model.
Method as described above, after the structure depth convolutional neural networks model, including：
Obtain diagnosis result corresponding to each electronic health record diagnosed；
For each electronic health record diagnosed, the diagnosis of the depth convolutional neural networks model output is obtained
Electronic health record corresponding to characteristic vector；
The characteristic vector of the electronic health record diagnosed is classified using the grader, diagnosed described in acquisition
Electronic health record each illness P；
Using preset algorithm to the P of each illness of electronic health record diagnosed and the electronic health record diagnosed
Corresponding diagnosis result is analyzed, according to the parameter of depth convolutional neural networks model and institute described in analysis result amendment
State the parameter of grader.
Method as described above, the preset algorithm are inverse iteration algorithm.
Method as described above, the grader are softmax graders.
Method as described above, the depth convolutional neural networks model include：Input layer, convolutional layer, pond layer, Quan Lian
Connect layer.
Method as described above, the convolutional layer include multiple various sizes of convolution kernels.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obtain substantially, or recognized by the practice of the present invention.
Brief description of the drawings
Of the invention above-mentioned and/or additional aspect and advantage will become from the following description of the accompanying drawings of embodiments
Substantially and it is readily appreciated that, wherein,
Fig. 1 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of one embodiment of the invention
Figure；
Fig. 2 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of further embodiment of this invention
Figure；
Fig. 3 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of another embodiment of the present invention
Figure；
Fig. 4 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of yet another embodiment of the invention
Figure；
Fig. 5 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of one embodiment of the invention
Figure；
Fig. 6 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of further embodiment of this invention
Figure；
Fig. 7 is the Organization Chart of the exemplary depth convolutional neural networks model of the embodiment of the present invention.
Embodiment
Embodiments of the invention are described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or the element with same or like function.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and is not considered as limiting the invention.
Below with reference to the accompanying drawings the method for describing the medical diagnosis based on depth convolutional neural networks of the embodiment of the present invention.
Deep learning correlation technique is briefly introduced herein.Deep learning (deep learning) is one of machine learning
Branch, be in machine learning it is a kind of based on to data carry out representative learning method, it attempt use comprising labyrinth or by
Multiple process layers that multiple nonlinear transformation is formed carry out higher level of abstraction to data.The benefit of deep learning be with non-supervisory formula or
Feature learning and layered characteristic the extraction highly effective algorithm of Semi-supervised obtain feature by hand to substitute.The target of representative learning is to seek
Seek more preferable method for expressing and create more preferable model to come from extensive these method for expressing of Unlabeled data learning.Expression side
Formula and loosely creates the information processing in similar nervous system and the understanding of communication pattern similar to the progress of Neuscience
On, such as neural coding, it is intended to definition pull neuron reaction between relation and the neuron in brain electrical activity it
Between relation.So far existing several deep learning frameworks, such as deep neural network, convolutional neural networks and depth confidence network and
Recurrent neural network has been employed the neck such as computer vision, speech recognition, natural language processing, audio identification and bioinformatics
Domain simultaneously obtains fabulous effect.
Convolutional neural networks (Convolutional Neural Network, CNN) are a kind of deep learning frameworks, and it is
A kind of feedforward neural network, it is made up of one or more convolutional layers and the full-mesh layer on top (corresponding classical neutral net),
Also include associated weights and pond layer (pooling layer) simultaneously.It is defeated that this structure enables convolutional neural networks to utilize
Enter the two-dimensional structure of data.Compare other depth, feedforward neural network, and the parameter that convolutional neural networks needs are estimated is less,
Make a kind of deep learning structure for having much attraction.A lot of research and application is early it was demonstrated that convolutional neural networks exist
There is very strong feature extraction, expression, semantic understanding ability on image and text, it does not need artificial design feature, but
Self-teaching is to various features from mass data.
Fig. 1 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of one embodiment of the invention
Figure.As shown in figure 1, the method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, including：
S101, obtain term vector matrix corresponding to electronic health record to be diagnosed.
Specifically, case history is original record of the patient in hospital diagnosis treatment overall process, and it includes homepage, course of disease note
Record, check assay, doctor's advice, operation record, nursing record etc..Electronic health record refers not only to static medical record information, in addition to
The related service of offer, it is the information of the relevant personal lifetime health state and health care behavior managed in a manner of electronic,
It is related to collection, storage, transmission, processing and all procedural informations utilized of patient information.It is pointed out that in electronic health record
Not only needed comprising medical vocabulary also comprising non-medical vocabulary such as many personal sensitive informations in the medical diagnosis of automation
Remove the distracters such as non-medical vocabulary, obtain the medical vocabulary in electronic health record to ensure the reliability of medical diagnosis.
In the present embodiment, first, the original electron case history for treating diagnosis carries out structuring and deleted such as personal quick
After feeling the non-medical vocabulary such as information, the plain text content of electronic health record to be diagnosed is obtained；Then, the electronic health record of diagnosis is treated
Plain text content carry out text identification, extract the medical vocabulary that electronic health record to be diagnosed includes；Followed by acquisition is each
The term vector of medical vocabulary, according to the term vector of each medical vocabulary with regard to word corresponding to getting electronic health record to be diagnosed to
Moment matrix.
In the present embodiment, the word that term vector can be understood as in language carries out expression-form, word in a manner of mathematicization
Vector can effectively solve the problems, such as " semantic gap ", effectively eliminate semantic ambiguity.The word of semantic similarity also has phase in vector space
Near distance, it so can effectively eliminate influence caused by synonymous different words in being write in electronic health record.
Briefly introduce how term vector matrix obtained according to term vector herein.For example, each electronic health record table of predefined
L*F term vector matrix is shown as, the first row represents the term vector of first word of text, and the second row represents the word of second word
Vector, by that analogy.For word length more than the electronic health record of L word, L word before interception；For electronics of the word length less than L word
Case history, follow-up term vector are filled with numeral 0.So each electronic health record is collectively expressed as L*F term vector matrix, is designated as M,
It can be expressed as：
Wherein,Bound symbol is represented, uses Xi:jThe matrix that i-th of term vector is formed to j-th of term vector is represented, per a line
Represent a term vector, X1:LThe matrix that the 1st term vector is formed to l-th term vector is represented, wherein, i, j, L, F are just whole
Number.
S102, by term vector Input matrix corresponding to electronic health record to be diagnosed to the depth convolutional Neural net built in advance
In network model, characteristic vector corresponding to electronic health record to be diagnosed is obtained.
In the present embodiment, term vector can effectively solve the problems, such as " semantic gap ", effectively eliminate semantic ambiguity, and then will
Term vector Input matrix corresponding to electronic health record to be diagnosed is acquired to the depth convolutional neural networks model built in advance
Characteristic vector can effectively eliminate influence caused by synonymous different words in writing in electronic health record.
Depth convolutional neural networks model in the present embodiment is trained by the electronic health record diagnosed of magnanimity
And obtain, these electronic health records diagnosed come from clinical history data, therefore the spy of depth convolutional neural networks model output
Sign vector has very strong practical guided significance.
In the present embodiment, and electronic health record need not be classified by disease, depth convolutional Neural is established by disease
Network model, a depth convolutional neural networks model can just examine the effect of multiple diseases in advance, be especially suitable for managing and safeguard, can
Autgmentability is also very strong.
S103, the characteristic vector for the electronic health record for being treated using grader diagnosis are classified, and obtain electronics to be diagnosed
The P of each illness corresponding to case history.
Specifically, the characteristic vector that the present embodiment is exported using grader to depth convolutional neural networks model carries out data
Excavate, the P for analyzing each illness in characteristic vector refers to for healthcare givers, realizes more preferable medical aided diagnosis effect
Fruit.Alternatively, grader is softmax graders, and the illness that softmax graders can preferably analyze each illness is general
Rate.
The method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, by obtaining electricity to be diagnosed
Term vector matrix corresponding to sub- case history；By term vector Input matrix corresponding to electronic health record to be diagnosed to the depth built in advance
In convolutional neural networks model, characteristic vector corresponding to electronic health record to be diagnosed is obtained；The electricity of diagnosis is treated using grader
The characteristic vector of sub- case history is classified, and obtains the P of each illness corresponding to electronic health record to be diagnosed.This method
Convolutional neural networks are applied into medical electronics case history text semantic to understand and carry out assisted medical diagnosis, can effectively overcome and be based on
Defect possessed by rule extraction and the method for matching.This method can effectively solve the problems, such as " semantic gap ", effectively eliminate electricity
The influence caused by synonymous different words etc. in writing of sub- case history；This method can be various for each section office's disease and difference is larger
Situation, it is not necessary to build corresponding rule and matching algorithm for each disease, unified model framework can be built, then with each
The historical data training pattern of individual disease, it is possible to very suitable as long as reaching the effect that a model can just examine multiple diseases in advance
Close management and safeguard, scalability is also very strong；This method does not need artificial design rule and feature, the institute that model learns
Have a feature and rule come from clinical a large amount of historical datas, be entirely to instruct clinical decision with clinical history data, compared to based on
The aided diagnosis method of rule has very strong practical guided significance.
Fig. 2 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of further embodiment of this invention
Figure.On the basis of above-described embodiment, optimization that the present embodiment " will obtain term vector matrix corresponding to electronic health record to be diagnosed "
For " electronic health record for treating diagnosis carries out at least one of information filtering, screening, participle, statistics operation, obtains and treats diagnosis disease
The each medical vocabulary gone through；Term vector corresponding to the medical vocabulary of case history to be diagnosed is obtained in default term vector database, its
In, preset the corresponding relation that medical vocabulary and term vector are preserved in term vector database；According to each electronic health record to be diagnosed
Medical vocabulary corresponding to term vector generation follow-up power-off son disease corresponding to term vector matrix.”
As shown in Fig. 2 the method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, including：
At least one of S201, the electronic health record progress information filtering for treating diagnosis, screening, participle, statistics operation, are obtained
Take each medical vocabulary of case history to be diagnosed.
Specifically, by carrying out the operation such as information filtering, screening, participle, statistics to electronic health record, medical vocabulary is protected
Stay, non-medical vocabulary is filtered out, be so advantageous to improve the processing speed of whole system.
S202, term vector corresponding to the medical vocabulary of case history to be diagnosed is obtained in default term vector database, wherein, in advance
If the corresponding relation of medical vocabulary and term vector is preserved in term vector database.
Specifically, preset term vector database save the medical vocabulary of magnanimity corresponding to term vector, also saving simultaneously
The corresponding relation of medical vocabulary and term vector.After each medical vocabulary of case history to be diagnosed is got, based on medical vocabulary
The medical word for carrying out case history to be diagnosed can be efficiently inquired in default term vector database with the corresponding relation of term vector
Term vector corresponding to remittance.The default term vector data for saving term vector corresponding to the medical vocabulary of magnanimity are utilized in the present embodiment
Storehouse, which is realized, efficiently obtains term vector corresponding to the medical vocabulary of case history to be diagnosed.
S203, term vector generation follow-up power-off son disease is right according to corresponding to the medical vocabulary of each electronic health record to be diagnosed
The term vector matrix answered.
Specifically, how to generate follow-up power-off son disease corresponding to term vector matrix, may refer to before to how according to word
Vector obtains the brief introduction of term vector matrix, will not be repeated here.
S204, by term vector Input matrix corresponding to electronic health record to be diagnosed to the depth convolutional Neural net built in advance
In network model, characteristic vector corresponding to electronic health record to be diagnosed is obtained.
S205, the characteristic vector for the electronic health record for being treated using grader diagnosis are classified, and obtain electronics to be diagnosed
The P of each illness corresponding to case history.
The implementation of step S204, S205 in the present embodiment reality with S102, S103 in above-described embodiment respectively
Existing mode is identical, will not be repeated here.
The method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, by being carried out to electronic health record
Information filtering is operated, and medical vocabulary is retained, non-medical vocabulary is filtered out, and is so advantageous to improve the processing speed of whole system
Degree；Efficiently obtain and treat by using the default term vector database realizing for saving term vector corresponding to the medical vocabulary of magnanimity
Diagnose term vector corresponding to the medical vocabulary of case history.
Fig. 3 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of another embodiment of the present invention
Figure.On the basis of above-described embodiment, the present embodiment optimizes explanation to establishing default term vector database.
As shown in figure 3, the method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, including：
Each medical vocabulary in S301, the medical dictionary of acquisition.
Specifically, medical dictionary is established in advance, and the medical vocabulary of magnanimity has been included in the medical dictionary.
S302, the medical vocabulary in medical dictionary is input in the Word2Vec models pre-established, obtains medical word
Term vector corresponding to remittance.
Word2Vec models are a kind of efficient tools that word is expressed as to real number value vector, can be reflected each word by training
It is vectorial (K is generally the hyper parameter in model) to penetrate into K dimension real numbers, and the similarity in vector space can be used for representing text
Similarity semantically.
S303, corresponding with medical vocabulary term vector formed into term vector sample, by term vector Sample preservation in default word
In vector data storehouse.
Specifically, preset term vector database save the medical vocabulary of magnanimity corresponding to term vector, also saving simultaneously
The corresponding relation of medical vocabulary and term vector.For example, after medical vocabulary is got, pair based on medical vocabulary and term vector
It should be related to efficiently to inquire in default term vector database and carry out term vector corresponding to medical vocabulary.
The method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, by obtaining in medical dictionary
Each medical vocabulary；Medical vocabulary in medical dictionary is input in the Word2Vec models pre-established, obtains medical treatment
Term vector corresponding to vocabulary；Term vector corresponding with medical vocabulary is formed into term vector sample, by term vector Sample preservation pre-
If in term vector database.Because medical dictionary has included the medical vocabulary of magnanimity, correspondingly, the default term vector number established
Term vector corresponding to the medical vocabulary of magnanimity is stored according to storehouse, and then is advantageous to the Effec-tive Function of whole system.In addition, utilize
Word2Vec models can efficiently get term vector corresponding to medical vocabulary.
Fig. 4 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of yet another embodiment of the invention
Figure.On the basis of above-described embodiment, the present embodiment illustrates to establishing medical dictionary.
As shown in figure 4, the method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, including：
S401, obtain multiple electronic health records diagnosed.
For example, the electronic health record diagnosed in the present embodiment can be carried from the medical information management system of hospital
Take the electronic health record of each section office's chief complaint of past few years.After the electronic health record diagnosed of magnanimity is got, to
The electronic health record of diagnosis carries out structuring, deletes the information such as personal sensitive information, obtains the plain text content of electronic health record.Its
In, multiple electronic health records diagnosed can be the electronic health record diagnosed of predetermined number.
S402, using Information Filtering Technology information filtering is carried out to each electronic health record for having diagnosed, obtain medical vocabulary
Set.
For example, the present embodiment first carries out semantic understanding to the plain text content of the electronic health record diagnosed of acquisition,
Obtain the medical vocabulary included of each electronic health record diagnosed.If electronic health record text is Chinese, it is necessary to be segmented
Operation, the segmentation word section of each electronic health record text is obtained, correspond to each word in English text.
In the present embodiment, the medical vocabulary of magnanimity can be obtained by analyzing the electronic health record diagnosed of magnanimity, correspondingly, by sea
The sample size of the medical lexical set of the medical vocabulary composition of amount is big.
The word frequency of each medical vocabulary in S403, the medical lexical set of statistics, each doctor is screened according to setting screening rule
Vocabulary is treated, medical dictionary is established according to the selection result.
For example, a medical dictionary is defined, is initially empty.Count the word of each medical vocabulary in medical lexical set
Frequently, setting screening rule can be that the medical vocabulary that word frequency is more than to given threshold is added in medical dictionary.The present embodiment passes through
The medical vocabulary of statistics is screened, the medical dictionary established more presses close to medical industry, authoritative higher.
The method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, by obtaining having examined for magnanimity
Disconnected electronic health record obtains the big medical lexical set of sample size, the medical vocabulary of statistics is screened, the medical treatment established
Dictionary more presses close to medical industry, authoritative higher.
Fig. 5 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of one embodiment of the invention
Figure.On the basis of above-described embodiment, the present embodiment illustrates to structure depth convolutional neural networks model.
As shown in figure 5, the method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, including：
S501, term vector matrix corresponding to multiple electronic health records diagnosed is obtained, the electronic health record diagnosed is corresponding
Term vector matrix as training sample.
For example, the electronic health record diagnosed can extract the past few years from the medical information management system of hospital
The electronic health record of each section office's chief complaint.After the electronic health record diagnosed of magnanimity is got, to the electronics disease diagnosed
Go through and carry out structuring, delete the information such as personal sensitive information, obtain the plain text content of electronic health record, row information of going forward side by side mistake
Filter, obtain medical vocabulary corresponding to each electronic health record diagnosed.Get diagnosed case history each medical vocabulary it
Afterwards, the corresponding relation based on medical vocabulary and term vector
Inquired in default term vector database and diagnosed term vector corresponding to the medical vocabulary of case history, last root
According to having diagnosed term vector matrix corresponding to the electronic health record that the generation of term vector corresponding to the medical vocabulary of case history diagnosed.
Specifically, for the depth convolutional neural networks model in the present embodiment, the electronic health record diagnosed can be seen
Into being training sample.In order to train the pre- depth convolutional neural networks model for examining multiple diseases, the training sample in the present embodiment
This is The more the better.
S502, training sample is trained, builds depth convolutional neural networks model.
Specifically, the depth convolutional neural networks model in the present embodiment is the mould based on deep neural network Structure Creating
Type.
For example, define all depth convolutional neural networks structures as shown in Figure 7, be divided into input layer, convolutional layer, pond layer,
Full articulamentum.
Wherein, input layer is text matrix M.For example, each training sample is input in input layer.
Wherein, the convolutional layer number of plies and every layer of convolution nuclear volume determine according to actual conditions, each convolution kernel width and should
The width of layer input matrix is identical, is set to F, is highly determined according to actual conditions, alternatively, convolutional layer includes multiple different sizes
Convolution kernel.If convolution kernel is highly h, convolution kernel is expressed as w ∈ RhF, R expression real numbers, then the convolution kernel is in Xi:jExtract at place
Feature ciFor：
ci=f (w*Xi:j+bi)
Wherein f is nonlinear function, biFor biasing.The convolution kernel slides from top to bottom on text matrix, obtains the layer
Characteristic spectrum c：
C=[c1,c1,…,cL-h+1]
It should be noted that said process describes the process that a convolution kernel produces a feature, realistic model exists
It is to have multiple convolutional layers during calculating, each layer has multiple convolution kernels, and the generation process of each feature is as above.
Wherein, pond layer is used for reducing the characteristic vector that convolutional layer exports by pond, while improves result and (be not easy out
Existing over-fitting).In the present embodiment, pondization can be that maximum pondization can also be average pond, if maximum pond, then then
It is
Wherein, characteristic vector caused by full articulamentum is expressed asThen export and be：
Y=w*z+bo
Wherein, output vector y dimension is N, and N is also the disease quantity of required prediction.W is the power of required study
Weight, boFor biasing.
The method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, obtain multiple electricity diagnosed
Term vector matrix corresponding to sub- case history, using term vector matrix corresponding to the electronic health record diagnosed as training sample, to training
Sample is trained, and builds depth convolutional neural networks model.This method is made by obtaining the electronic health record diagnosed of magnanimity
For historical data, constructed depth convolutional neural networks model can examine the effect of multiple diseases in advance, be especially suitable for management and
Safeguard, scalability is also very strong；This method does not need artificial design rule and feature, all features that model learns and
Rule comes from clinical a large amount of historical datas, is entirely to instruct clinical decision with clinical history data, auxiliary compared to rule-based
Auxiliary diagnosis method has very strong practical guided significance.
Fig. 6 is the flow signal of the method for the medical diagnosis based on depth convolutional neural networks of further embodiment of this invention
Figure.On the basis of above-described embodiment, the present embodiment optimizes explanation to depth convolutional neural networks model and grader.
As shown in fig. 6, the method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, including：
S601, term vector matrix corresponding to multiple electronic health records diagnosed is obtained, the electronic health record diagnosed is corresponding
Term vector matrix as training sample.
S602, training sample is trained, builds depth convolutional neural networks model.
The implementation of step S601, S602 in the present embodiment reality with S501, S502 in above-described embodiment respectively
Existing mode is identical, will not be repeated here.
S603, obtain diagnosis result corresponding to each electronic health record diagnosed.
In the present embodiment, diagnosis result can be understood as doctor according to electronics corresponding to the electronic health record diagnosed
The medical diagnosis that various medical indexs in case history are made.
S604, the electronic health record for each having diagnosed, obtain having diagnosed for depth convolutional neural networks models output
Characteristic vector corresponding to electronic health record.
In the present embodiment, depth convolutional neural networks model is the electronic health record diagnosed using magnanimity as training sample
It is trained and obtains, in each training sample input depth convolutional neural networks model, depth convolutional neural networks model is defeated
Go out corresponding training result.
S605, classified using the characteristic vector of electronic health record of the grader to having diagnosed, obtain the electronics diagnosed
The P of each illness of case history.
Specifically, so that grader is softmax graders as an example, referring to Fig. 7, the complete of depth convolutional neural networks model connects
Softmax graders can be connected by connecing layer, and softmax graders divide the characteristic vector of depth convolutional neural networks model
Class, obtain the P of each illness：
Wherein, piRepresent the P of i-th of illness of the electronic health record diagnosed, yiY i-th of element is represented, i.e.,
Characteristic vector, y corresponding to i-th of illness in electronic health recordjRepresent y j-th of element, i.e., j-th of illness pair in electronic health record
The characteristic vector answered.
S606, using preset algorithm to the P of each illness of electronic health record diagnosed and the electronics diagnosed
Diagnosis result is analyzed corresponding to case history, according to the parameter of analysis result Corrected Depth convolutional neural networks model and is divided
The parameter of class device.
Specifically, the present embodiment in the training process, passes through such as backpropagation (Back Propagation, BP) algorithm
Etc. the parameter of preset algorithm renewal depth convolutional neural networks model and the parameter of grader.
For example, the object function of the output of whole network is expressed as：
It should be noted that whole network can be understood as the net from depth convolutional neural networks model and grader composition
Network.Wherein, the output result of P presentation classes device, each PiElement representation ill disease be i-th illness P.T
For actual value, i.e. T is the diagnosis result of electronic health record, i.e. corresponding t-th of the illness of such as electronic health record, then t in T vectors
Individual element value is 1, and its residual value is 0.Num represents the sample size trained every time.W represents the ginseng of depth convolutional neural networks model
Number.Training process is by BP (Back Propagation, backpropagation) algorithmic minimizing LOSS functions, until network convergence,
LOSS no longer declines, and now training is completed, and retains all parameters in whole network.
It is pointed out that P generation of the T vectors according to each illness of the electronic health record diagnosed.
For example, in test phase, all parameters in the whole network retained is constantly read and are used for updating depth volume
The parameter of product neural network model and the parameter of grader.Electronic health record is inputted into depth convolutional neural networks model, and profit
The output result corresponding to grader obtains the electronic health record is P vectors, and P vectors are the ill disease prediction knot of institute of the electronic health record
Fruit, if j-th of element value is maximum in P vectors, the ill disease of the most probable institute of prediction patient is j-th of illness.
The method for the medical diagnosis based on depth convolutional neural networks that the present embodiment provides, using grader to having diagnosed
The characteristic vector of electronic health record classified, the P of each illness of the electronic health record diagnosed is obtained, using anti-
To iterative algorithm to doctor corresponding to the P of each illness of electronic health record diagnosed and the electronic health record diagnosed
Diagnostic result is analyzed, and according to the parameter of analysis result Corrected Depth convolutional neural networks model and the parameter of grader, is entered
One step enables constructed depth convolutional neural networks model to examine the effect of multiple diseases in advance, is especially suitable for managing and safeguards,
Scalability is also very strong；This method does not need artificial design rule and feature, all features and rule that model learns
Come from clinical a large amount of historical datas, be entirely to instruct clinical decision with clinical history data, examined compared to rule-based auxiliary
Disconnected method has very strong practical guided significance.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or the spy for combining the embodiment or example description
Point is contained at least one embodiment or example of the present invention.In this manual, to the schematic representation of above-mentioned term not
Identical embodiment or example must be directed to.Moreover, specific features, structure, material or the feature of description can be with office
Combined in an appropriate manner in one or more embodiments or example.In addition, in the case of not conflicting, the skill of this area
Art personnel can be tied the different embodiments or example and the feature of different embodiments or example described in this specification
Close and combine.
In addition, term " first ", " second " are only used for describing purpose, and it is not intended that instruction or hint relative importance
Or the implicit quantity for indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the invention, " multiple " are meant that at least two, such as two, three
It is individual etc., unless otherwise specifically defined.
Any process or method described otherwise above description in flow chart or herein is construed as, and represents to include
Module, fragment or the portion of the code of the executable instruction of one or more the step of being used to realize custom logic function or process
Point, and the scope of the preferred embodiment of the present invention includes other realization, wherein can not press shown or discuss suitable
Sequence, including according to involved function by it is basic simultaneously in the way of or in the opposite order, carry out perform function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
Expression or logic and/or step described otherwise above herein in flow charts, for example, being considered use
In the order list for the executable instruction for realizing logic function, may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (such as computer based system including the system of processor or other can be held from instruction
The system of row system, device or equipment instruction fetch and execute instruction) use, or combine these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium " can any can be included, store, communicate, propagate or pass
Defeated program is for instruction execution system, device or equipment or the dress used with reference to these instruction execution systems, device or equipment
Put.The more specifically example (non-exhaustive list) of computer-readable medium includes following：Electricity with one or more wiring
Connecting portion (electronic installation), portable computer diskette box (magnetic device), random access memory (RAM), read-only storage
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device, and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, computer-readable medium can even is that can the paper of print routine thereon or other suitable be situated between
Matter, because can then enter edlin, interpretation or if necessary with other for example by carrying out optical scanner to paper or other media
Suitable method is handled electronically to obtain program, is then stored in computer storage.
It should be appreciated that each several part of the present invention can be realized with hardware, software, firmware or combinations thereof.Above-mentioned
In embodiment, software that multiple steps or method can be performed in memory and by suitable instruction execution system with storage
Or firmware is realized.Such as, if realized with hardware with another embodiment, following skill well known in the art can be used
Any one of art or their combination are realized：With the logic gates for realizing logic function to data-signal from
Logic circuit is dissipated, the application specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can compile
Journey gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method carries
Suddenly be can by program come instruct correlation hardware complete, program can be stored in a kind of computer-readable recording medium
In, the program upon execution, including one or a combination set of the step of embodiment of the method.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing module, can also
That unit is individually physically present, can also two or more units be integrated in a module.Above-mentioned integrated mould
Block can both be realized in the form of hardware, can also be realized in the form of software function module.If integrated module with
The form of software function module realize and be used as independent production marketing or in use, can also be stored in one it is computer-readable
Take in storage medium.
Storage medium mentioned above can be read-only storage, disk or CD etc..Although have been shown and retouch above
Embodiments of the invention are stated, it is to be understood that above-described embodiment is exemplary, it is impossible to be interpreted as the limit to the present invention
System, one of ordinary skill in the art can be changed to above-described embodiment, change, replace and become within the scope of the invention
Type.
Claims (10)
- A kind of 1. method of the medical diagnosis based on depth convolutional neural networks, it is characterised in that including：Obtain term vector matrix corresponding to electronic health record to be diagnosed；By term vector Input matrix corresponding to the electronic health record to be diagnosed to the depth convolutional neural networks mould built in advance In type, characteristic vector corresponding to electronic health record to be diagnosed is obtained；The characteristic vector of the electronic health record to be diagnosed is classified using grader, electronics to be diagnosed disease described in acquisition The P of each illness corresponding to going through.
- 2. the method as described in claim 1, it is characterised in that described to obtain term vector square corresponding to electronic health record to be diagnosed Battle array, including：The electronic health record for treating diagnosis carries out at least one of information filtering, screening, participle, statistics operation, obtains and waits to diagnose Each medical vocabulary of case history；Term vector corresponding to the medical vocabulary of case history to be diagnosed described in being obtained in the default term vector database, wherein, institute State the corresponding relation that medical vocabulary and term vector are preserved in default term vector database；According to corresponding to term vector generation follow-up power-off son disease corresponding to the medical vocabulary of each electronic health record to be diagnosed Term vector matrix.
- 3. method as claimed in claim 2, it is characterised in that obtain term vector corresponding to electronic health record to be diagnosed described Before matrix, including：Obtain each medical vocabulary in medical dictionary；Medical vocabulary in the medical dictionary is input in the Word2Vec models pre-established, obtains the medical vocabulary Corresponding term vector；Term vector corresponding with the medical vocabulary is formed into term vector sample, by the term vector Sample preservation default word to Measure in database.
- 4. method as claimed in claim 3, it is characterised in that each medical vocabulary obtained in medical dictionary it Before, including：Obtain multiple electronic health records diagnosed；Information filtering is carried out to each electronic health record diagnosed using Information Filtering Technology, obtains medical lexical set；The word frequency of each medical vocabulary in the medical lexical set is counted, each medical word is screened according to setting screening rule Converge, the medical dictionary is established according to the selection result.
- 5. the method as described in claim 1, it is characterised in that obtain term vector corresponding to electronic health record to be diagnosed described Before matrix, including：Obtain term vector matrix corresponding to multiple electronic health records diagnosed, by word corresponding to the electronic health record diagnosed to Moment matrix is as training sample；The training sample is trained, builds the depth convolutional neural networks model.
- 6. method as claimed in claim 5, it is characterised in that the structure depth convolutional neural networks model it Afterwards, including：Obtain diagnosis result corresponding to each electronic health record diagnosed；For each electronic health record diagnosed, the electricity diagnosed that the depth convolutional neural networks model exports is obtained Characteristic vector corresponding to sub- case history；The characteristic vector of the electronic health record diagnosed is classified using the grader, the electricity diagnosed described in acquisition The P of each illness of sub- case history；It is corresponding with the electronic health record diagnosed to the P of each illness of the electronic health record diagnosed using preset algorithm Diagnosis result analyzed, according to the parameter of depth convolutional neural networks model described in analysis result amendment and described point The parameter of class device.
- 7. method as claimed in claim 6, it is characterised in that the preset algorithm is back-propagation algorithm.
- 8. the method as described in any one of claim 1 to 7, it is characterised in that the grader is softmax graders.
- 9. the method as described in any one of claim 1 to 7, it is characterised in that the depth convolutional neural networks model includes： Input layer, convolutional layer, pond layer, full articulamentum.
- 10. method as claimed in claim 9, it is characterised in that the convolutional layer includes multiple various sizes of convolution kernels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711002335.7A CN107863147B (en) | 2017-10-24 | 2017-10-24 | Medical diagnosis method based on deep convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711002335.7A CN107863147B (en) | 2017-10-24 | 2017-10-24 | Medical diagnosis method based on deep convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107863147A true CN107863147A (en) | 2018-03-30 |
| CN107863147B CN107863147B (en) | 2021-03-16 |
Family
ID=61696344
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711002335.7A Active CN107863147B (en) | 2017-10-24 | 2017-10-24 | Medical diagnosis method based on deep convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107863147B (en) |
Cited By (39)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108766561A (en) * | 2018-05-31 | 2018-11-06 | 平安医疗科技有限公司 | Illness information processing method, device, computer equipment and storage medium |
| CN108877923A (en) * | 2018-05-03 | 2018-11-23 | 华南理工大学 | A method of the tongue fur based on deep learning generates prescriptions of traditional Chinese medicine |
| CN108877929A (en) * | 2018-05-31 | 2018-11-23 | 平安医疗科技有限公司 | It is neural network based to intend examining recommendation process method, apparatus and computer equipment |
| CN109003677A (en) * | 2018-06-11 | 2018-12-14 | 清华大学 | Medical record data structured analysis processing method |
| CN109192299A (en) * | 2018-08-13 | 2019-01-11 | 中国科学院计算技术研究所 | A kind of medical analysis auxiliary system based on convolutional neural networks |
| CN109243618A (en) * | 2018-09-12 | 2019-01-18 | 腾讯科技（深圳）有限公司 | Construction method, disease label construction method and the smart machine of medical model |
| CN109378066A (en) * | 2018-12-20 | 2019-02-22 | 翼健(上海)信息科技有限公司 | A kind of control method and control device for realizing disease forecasting based on feature vector |
| CN109390058A (en) * | 2018-09-28 | 2019-02-26 | 湖南智腾安控科技有限公司 | A kind of method for building up of case history Computer Aided Analysis System and the system |
| CN109411062A (en) * | 2018-09-30 | 2019-03-01 | 天津科技大学 | A method of diabetes and its complication are judged automatically using physiological parameter based on depth learning technology |
| CN109493931A (en) * | 2018-10-25 | 2019-03-19 | 平安科技（深圳）有限公司 | A kind of coding method of patient file, server and computer readable storage medium |
| CN109599185A (en) * | 2018-11-14 | 2019-04-09 | 金色熊猫有限公司 | Disease data processing method, device, electronic equipment and computer-readable medium |
| CN109698016A (en) * | 2018-12-11 | 2019-04-30 | 中国科学院深圳先进技术研究院 | Disease automatic coding and device |
| CN109801706A (en) * | 2018-12-12 | 2019-05-24 | 清华大学 | The cognitive method and device of psychological pressure problem |
| CN109920501A (en) * | 2019-01-24 | 2019-06-21 | 西安交通大学 | Electronic health record classification method and system based on convolutional neural networks and Active Learning |
| CN109919243A (en) * | 2019-03-15 | 2019-06-21 | 天津拾起卖科技有限公司 | A kind of scrap iron and steel type automatic identifying method and device based on CNN |
| CN109949927A (en) * | 2019-02-18 | 2019-06-28 | 四川拾智联兴科技有限公司 | A kind of intelligent diagnosing method and its system based on deep neural network |
| CN109949929A (en) * | 2019-03-19 | 2019-06-28 | 挂号网（杭州）科技有限公司 | A kind of assistant diagnosis system based on the extensive case history of deep learning |
| CN110021439A (en) * | 2019-03-07 | 2019-07-16 | 平安科技（深圳）有限公司 | Medical data classification method, device and computer equipment based on machine learning |
| CN110136788A (en) * | 2019-05-14 | 2019-08-16 | 清华大学 | It is a kind of based on the case history quality detecting method, device, equipment and the storage medium that detect automatically |
| CN110349668A (en) * | 2019-07-11 | 2019-10-18 | 四川大学 | A kind of therapeutic scheme aid decision-making method and its system based on BP neural network |
| CN110471941A (en) * | 2019-08-12 | 2019-11-19 | 贵州医渡云技术有限公司 | It is automatically positioned the method, apparatus and electronic equipment of judgment basis |
| CN110755045A (en) * | 2019-10-30 | 2020-02-07 | 湖南财政经济学院 | Skin disease comprehensive data analysis and diagnosis auxiliary system and information processing method |
| CN110911009A (en) * | 2019-11-14 | 2020-03-24 | 南京医科大学 | Clinical diagnosis aid decision-making system and medical knowledge map accumulation method |
| CN110929511A (en) * | 2018-09-04 | 2020-03-27 | 清华大学 | Intelligent matching method for personalized traditional Chinese medicine diagnosis and treatment information and traditional Chinese medicine information based on semantic similarity |
| CN111081379A (en) * | 2019-12-02 | 2020-04-28 | 中南大学 | Disease probability decision method and system |
| CN111199797A (en) * | 2019-12-31 | 2020-05-26 | 中国中医科学院中医药信息研究所 | Auxiliary diagnosis model establishing and auxiliary diagnosis method and device |
| CN111199801A (en) * | 2018-11-19 | 2020-05-26 | 零氪医疗智能科技(广州)有限公司 | Construction method and application of model for identifying disease types of medical records |
| CN111259116A (en) * | 2020-01-16 | 2020-06-09 | 北京珞安科技有限责任公司 | Sensitive file detection method based on convolutional neural network |
| CN111462895A (en) * | 2020-03-30 | 2020-07-28 | 安徽科大讯飞医疗信息技术有限公司 | Auxiliary diagnosis method and system |
| CN111599462A (en) * | 2020-05-09 | 2020-08-28 | 吾征智能技术(北京)有限公司 | Intelligent body abnormal smell screening system based on cognitive learning |
| CN111599463A (en) * | 2020-05-09 | 2020-08-28 | 吾征智能技术(北京)有限公司 | Intelligent auxiliary diagnosis system based on sound cognition model |
| CN112001066A (en) * | 2020-07-30 | 2020-11-27 | 四川大学 | Deep learning-based method for calculating limit transmission capacity |
| CN112259229A (en) * | 2020-06-10 | 2021-01-22 | 北京小白世纪网络科技有限公司 | Method, system and equipment for realizing traditional Chinese medicine diagnosis and treatment based on neural network |
| CN112347776A (en) * | 2019-08-09 | 2021-02-09 | 金色熊猫有限公司 | Medical data processing method and device, storage medium and electronic equipment |
| WO2021027284A1 (en) * | 2019-08-12 | 2021-02-18 | 北京国双科技有限公司 | Method and device for determining comment nature of text |
| CN112466462A (en) * | 2020-11-26 | 2021-03-09 | 华侨大学 | EMR information association and evolution method based on deep learning of image |
| CN112951432A (en) * | 2021-03-30 | 2021-06-11 | 太原理工大学 | Method and system for predicting heart failure complications based on CNN and electronic medical record, early warning equipment and storage medium |
| CN113658688A (en) * | 2021-08-23 | 2021-11-16 | 上海赴源科技服务有限公司 | Clinical decision support method based on word segmentation-free deep learning |
| CN109493931B (en) * | 2018-10-25 | 2024-06-04 | 平安科技（深圳）有限公司 | Medical record file encoding method, server and computer readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9430829B2 (en) * | 2014-01-30 | 2016-08-30 | Case Western Reserve University | Automatic detection of mitosis using handcrafted and convolutional neural network features |
| CN106446526A (en) * | 2016-08-31 | 2017-02-22 | 北京千安哲信息技术有限公司 | Electronic medical record entity relation extraction method and apparatus |
| CN106778014A (en) * | 2016-12-29 | 2017-05-31 | 浙江大学 | A kind of risk Forecasting Methodology based on Recognition with Recurrent Neural Network |
| CN106874643A (en) * | 2016-12-27 | 2017-06-20 | 中国科学院自动化研究所 | Build the method and system that knowledge base realizes assisting in diagnosis and treatment automatically based on term vector |
| CN107220506A (en) * | 2017-06-05 | 2017-09-29 | 东华大学 | Breast cancer risk assessment analysis system based on depth convolutional neural networks |
| 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 |
-
2017
- 2017-10-24 CN CN201711002335.7A patent/CN107863147B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9430829B2 (en) * | 2014-01-30 | 2016-08-30 | Case Western Reserve University | Automatic detection of mitosis using handcrafted and convolutional neural network features |
| CN106446526A (en) * | 2016-08-31 | 2017-02-22 | 北京千安哲信息技术有限公司 | Electronic medical record entity relation extraction method and apparatus |
| CN106874643A (en) * | 2016-12-27 | 2017-06-20 | 中国科学院自动化研究所 | Build the method and system that knowledge base realizes assisting in diagnosis and treatment automatically based on term vector |
| CN106778014A (en) * | 2016-12-29 | 2017-05-31 | 浙江大学 | A kind of risk Forecasting Methodology based on Recognition with Recurrent Neural Network |
| CN107220506A (en) * | 2017-06-05 | 2017-09-29 | 东华大学 | Breast cancer risk assessment analysis system based on depth convolutional neural networks |
| 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 |
Non-Patent Citations (2)
| Title |
|---|
| XISHUANG DONG 等: "A multiclass classification method based on deep learning for named entity recognition in electronic medical records", 《2016 NEW YORK SCIENTIFIC DATA SUMMIT (NYSDS)》 * |
| 陈凯: "基于患者个人信息的智能预诊方法", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (54)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108877923A (en) * | 2018-05-03 | 2018-11-23 | 华南理工大学 | A method of the tongue fur based on deep learning generates prescriptions of traditional Chinese medicine |
| CN108766561B (en) * | 2018-05-31 | 2023-12-08 | 平安医疗科技有限公司 | Method, apparatus, computer device and storage medium for processing disease information |
| CN108877929A (en) * | 2018-05-31 | 2018-11-23 | 平安医疗科技有限公司 | It is neural network based to intend examining recommendation process method, apparatus and computer equipment |
| CN108766561A (en) * | 2018-05-31 | 2018-11-06 | 平安医疗科技有限公司 | Illness information processing method, device, computer equipment and storage medium |
| CN109003677A (en) * | 2018-06-11 | 2018-12-14 | 清华大学 | Medical record data structured analysis processing method |
| CN109192299A (en) * | 2018-08-13 | 2019-01-11 | 中国科学院计算技术研究所 | A kind of medical analysis auxiliary system based on convolutional neural networks |
| CN110929511B (en) * | 2018-09-04 | 2021-12-17 | 清华大学 | Intelligent matching method for personalized traditional Chinese medicine diagnosis and treatment information and traditional Chinese medicine information based on semantic similarity |
| CN110929511A (en) * | 2018-09-04 | 2020-03-27 | 清华大学 | Intelligent matching method for personalized traditional Chinese medicine diagnosis and treatment information and traditional Chinese medicine information based on semantic similarity |
| CN109243618B (en) * | 2018-09-12 | 2020-06-16 | 腾讯科技（深圳）有限公司 | Medical model construction method, disease label construction method and intelligent device |
| CN109243618A (en) * | 2018-09-12 | 2019-01-18 | 腾讯科技（深圳）有限公司 | Construction method, disease label construction method and the smart machine of medical model |
| CN109390058A (en) * | 2018-09-28 | 2019-02-26 | 湖南智腾安控科技有限公司 | A kind of method for building up of case history Computer Aided Analysis System and the system |
| CN109411062A (en) * | 2018-09-30 | 2019-03-01 | 天津科技大学 | A method of diabetes and its complication are judged automatically using physiological parameter based on depth learning technology |
| CN109493931A (en) * | 2018-10-25 | 2019-03-19 | 平安科技（深圳）有限公司 | A kind of coding method of patient file, server and computer readable storage medium |
| CN109493931B (en) * | 2018-10-25 | 2024-06-04 | 平安科技（深圳）有限公司 | Medical record file encoding method, server and computer readable storage medium |
| CN109599185A (en) * | 2018-11-14 | 2019-04-09 | 金色熊猫有限公司 | Disease data processing method, device, electronic equipment and computer-readable medium |
| CN111199801B (en) * | 2018-11-19 | 2023-08-08 | 零氪医疗智能科技(广州)有限公司 | Construction method and application of model for identifying disease types of medical records |
| CN111199801A (en) * | 2018-11-19 | 2020-05-26 | 零氪医疗智能科技(广州)有限公司 | Construction method and application of model for identifying disease types of medical records |
| CN109698016A (en) * | 2018-12-11 | 2019-04-30 | 中国科学院深圳先进技术研究院 | Disease automatic coding and device |
| CN109801706A (en) * | 2018-12-12 | 2019-05-24 | 清华大学 | The cognitive method and device of psychological pressure problem |
| CN109378066A (en) * | 2018-12-20 | 2019-02-22 | 翼健(上海)信息科技有限公司 | A kind of control method and control device for realizing disease forecasting based on feature vector |
| CN109920501A (en) * | 2019-01-24 | 2019-06-21 | 西安交通大学 | Electronic health record classification method and system based on convolutional neural networks and Active Learning |
| CN109920501B (en) * | 2019-01-24 | 2021-04-20 | 西安交通大学 | Electronic medical record classification method and system based on convolutional neural network and active learning |
| CN109949927A (en) * | 2019-02-18 | 2019-06-28 | 四川拾智联兴科技有限公司 | A kind of intelligent diagnosing method and its system based on deep neural network |
| JP7162726B2 (en) | 2019-03-07 | 2022-10-28 | 平安科技（深▲せん▼）有限公司 | Medical data classification method, apparatus, computer device and storage medium based on machine learning |
| JP2021532499A (en) * | 2019-03-07 | 2021-11-25 | 平安科技(深▲せん▼)有限公司Ping An Technology (Shenzhen) Co., Ltd. | Machine learning-based medical data classification methods, devices, computer devices and storage media |
| CN110021439A (en) * | 2019-03-07 | 2019-07-16 | 平安科技（深圳）有限公司 | Medical data classification method, device and computer equipment based on machine learning |
| CN109919243A (en) * | 2019-03-15 | 2019-06-21 | 天津拾起卖科技有限公司 | A kind of scrap iron and steel type automatic identifying method and device based on CNN |
| CN109949929A (en) * | 2019-03-19 | 2019-06-28 | 挂号网（杭州）科技有限公司 | A kind of assistant diagnosis system based on the extensive case history of deep learning |
| CN110136788B (en) * | 2019-05-14 | 2021-08-17 | 清华大学 | Medical record quality inspection method, device, equipment and storage medium based on automatic detection |
| CN110136788A (en) * | 2019-05-14 | 2019-08-16 | 清华大学 | It is a kind of based on the case history quality detecting method, device, equipment and the storage medium that detect automatically |
| CN110349668A (en) * | 2019-07-11 | 2019-10-18 | 四川大学 | A kind of therapeutic scheme aid decision-making method and its system based on BP neural network |
| CN112347776B (en) * | 2019-08-09 | 2024-05-24 | 金色熊猫有限公司 | Medical data processing method and device, storage medium and electronic equipment |
| CN112347776A (en) * | 2019-08-09 | 2021-02-09 | 金色熊猫有限公司 | Medical data processing method and device, storage medium and electronic equipment |
| WO2021027284A1 (en) * | 2019-08-12 | 2021-02-18 | 北京国双科技有限公司 | Method and device for determining comment nature of text |
| CN110471941A (en) * | 2019-08-12 | 2019-11-19 | 贵州医渡云技术有限公司 | It is automatically positioned the method, apparatus and electronic equipment of judgment basis |
| CN110755045B (en) * | 2019-10-30 | 2022-06-07 | 湖南财政经济学院 | Skin disease comprehensive data analysis and diagnosis auxiliary system |
| CN110755045A (en) * | 2019-10-30 | 2020-02-07 | 湖南财政经济学院 | Skin disease comprehensive data analysis and diagnosis auxiliary system and information processing method |
| CN110911009A (en) * | 2019-11-14 | 2020-03-24 | 南京医科大学 | Clinical diagnosis aid decision-making system and medical knowledge map accumulation method |
| CN111081379A (en) * | 2019-12-02 | 2020-04-28 | 中南大学 | Disease probability decision method and system |
| CN111081379B (en) * | 2019-12-02 | 2023-05-16 | 中南大学 | Disease probability decision method and system thereof |
| CN111199797A (en) * | 2019-12-31 | 2020-05-26 | 中国中医科学院中医药信息研究所 | Auxiliary diagnosis model establishing and auxiliary diagnosis method and device |
| CN111259116A (en) * | 2020-01-16 | 2020-06-09 | 北京珞安科技有限责任公司 | Sensitive file detection method based on convolutional neural network |
| CN111462895A (en) * | 2020-03-30 | 2020-07-28 | 安徽科大讯飞医疗信息技术有限公司 | Auxiliary diagnosis method and system |
| CN111462895B (en) * | 2020-03-30 | 2024-04-05 | 讯飞医疗科技股份有限公司 | Auxiliary diagnosis method and system |
| CN111599463A (en) * | 2020-05-09 | 2020-08-28 | 吾征智能技术(北京)有限公司 | Intelligent auxiliary diagnosis system based on sound cognition model |
| CN111599462A (en) * | 2020-05-09 | 2020-08-28 | 吾征智能技术(北京)有限公司 | Intelligent body abnormal smell screening system based on cognitive learning |
| CN111599462B (en) * | 2020-05-09 | 2023-11-21 | 吾征智能技术(北京)有限公司 | Intelligent body abnormal odor screening system based on cognitive learning |
| CN112259229A (en) * | 2020-06-10 | 2021-01-22 | 北京小白世纪网络科技有限公司 | Method, system and equipment for realizing traditional Chinese medicine diagnosis and treatment based on neural network |
| CN112001066A (en) * | 2020-07-30 | 2020-11-27 | 四川大学 | Deep learning-based method for calculating limit transmission capacity |
| CN112001066B (en) * | 2020-07-30 | 2022-11-04 | 四川大学 | Deep learning-based method for calculating limit transmission capacity |
| CN112466462A (en) * | 2020-11-26 | 2021-03-09 | 华侨大学 | EMR information association and evolution method based on deep learning of image |
| CN112951432A (en) * | 2021-03-30 | 2021-06-11 | 太原理工大学 | Method and system for predicting heart failure complications based on CNN and electronic medical record, early warning equipment and storage medium |
| CN113658688B (en) * | 2021-08-23 | 2024-02-13 | 上海赴源科技服务有限公司 | Clinical decision support method based on word segmentation-free deep learning |
| CN113658688A (en) * | 2021-08-23 | 2021-11-16 | 上海赴源科技服务有限公司 | Clinical decision support method based on word segmentation-free deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107863147B (en) | 2021-03-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107863147A (en) | The method of medical diagnosis based on depth convolutional neural networks | |
| Yang et al. | Explaining therapy predictions with layer-wise relevance propagation in neural networks | |
| EP3734604A1 (en) | Method and system for supporting medical decision making | |
| CN113421652A (en) | Method for analyzing medical data, method for training model and analyzer | |
| Almadhoun et al. | Detection of brain tumor using deep learning | |
| CN109559300A (en) | Image processing method, electronic equipment and computer readable storage medium | |
| CN108563626A (en) | Medical text name entity recognition method and device | |
| CN109192299A (en) | A kind of medical analysis auxiliary system based on convolutional neural networks | |
| Levakov et al. | Mapping individual differences across brain network structure to function and behavior with connectome embedding | |
| CN113035362A (en) | Medical prediction method and system based on semantic graph network | |
| Srivani et al. | Cognitive computing technological trends and future research directions in healthcare—A systematic literature review | |
| Elshoky et al. | Comparing automated and non‐automated machine learning for autism spectrum disorders classification using facial images | |
| Pakzad et al. | CIRCLe: Color invariant representation learning for unbiased classification of skin lesions | |
| Lucieri et al. | Achievements and challenges in explaining deep learning based computer-aided diagnosis systems | |
| Hespel et al. | Artificial intelligence 101 for veterinary diagnostic imaging | |
| Sathish et al. | IoT based Mobile App for Skin Cancer Detection using Transfer Learning | |
| Songram et al. | A study of features affecting on stroke prediction using machine learning | |
| CN114300122A (en) | Crop disease diagnosis method and system based on electronic medical record | |
| CN107910066A (en) | Case history appraisal procedure, device, electronic equipment and storage medium | |
| Gnanasankaran et al. | Amalgamation of deep learning in healthcare systems | |
| binti Omar et al. | Personalized clinical pathway for heart failure management | |
| Grimm et al. | Modeling latent attention within neural networks | |
| Abut et al. | Deep Neural Networks and Applications in Medical Research | |
| Shyamala et al. | Machine Learning Algorithms for Healthcare Data Analytics | |
| Shellshear et al. | Machine learning, deep learning and neural networks |
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 |