CN106055896A - TCM (Traditional Chinese Medicine) in-vitro diagnosis method and device - Google Patents
TCM (Traditional Chinese Medicine) in-vitro diagnosis method and device Download PDFInfo
- Publication number
- CN106055896A CN106055896A CN201610377994.8A CN201610377994A CN106055896A CN 106055896 A CN106055896 A CN 106055896A CN 201610377994 A CN201610377994 A CN 201610377994A CN 106055896 A CN106055896 A CN 106055896A
- Authority
- CN
- China
- Prior art keywords
- vitro detection
- detection information
- rarefaction representation
- sample
- characteristic 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
Abstract
The invention discloses a TCM (Traditional Chinese Medicine) in-vitro diagnosis method. The TCM in-vitro diagnosis method comprises the following steps of obtaining a test sample, wherein the test sample is composed of obtained K in-vitro detection information of users; and the K is a positive integer, which is greater than or equal to 2; performing sparse representation of the K in-vitro detection information so as to obtain a sparse representation coefficient corresponding to each piece of in-vitro detection information; and diagnosing the test sample according to the sparse representation coefficient, a pre-set training sample and a sparse classifier. The invention further discloses a TCM in-vitro diagnosis device. According to the TCM in-vitro diagnosis method and device disclosed by the invention, fusion analysis of multi-aspect characteristics of human bodies is carried out, so that the automatic diagnosis purpose is realized.
Description
Technical field
The present invention relates to tcm diagnosis field, particularly relate to traditional Chinese medical science in-vitro diagnosis method and device.
Background technology
Along with the raising of people's living standard, the research of human health problems is increasingly extensive.Wherein Traditional Chinese Medicine has because of it
Standby non-invasive and convenience, is widely used in fields such as human body constitution identification and medicals diagnosis on disease.The thing followed is each
Kind traditional Chinese medical science equipment have also been obtained and develops widely.
But, the feature in a certain respect that existing diagnostic method based on the traditional Chinese medical science and equipment often only considered in the traditional Chinese medical science is entered
Row diagnosis, diagnoses as obtained picture of the tongue or pulse condition information.But during reality diagnoses, need to combine the letter of multiple aspect
Breath, if picture of the tongue, face are as, the information such as pulse condition, abnormal smells from the patient, carries out diagnosing diagnostic result just can be made the most accurate, therefore obtain human body
It is current problem demanding prompt solution that many features convergence analysis carry out diagnosis.
Summary of the invention
Present invention is primarily targeted at a kind of traditional Chinese medical science in-vitro diagnosis method and device of offer, it is intended to by by human body in many ways
Region feature carries out convergence analysis and realizes automatic diagnostic purpose.
For achieving the above object, a kind of traditional Chinese medical science in-vitro diagnosis method that the present invention provides comprises the following steps:
Obtaining test sample, described test sample is made up of the K kind vitro detection information of the user got, and K is the most whole
Count and more than or equal to 2；
Described K kind vitro detection information is carried out rarefaction representation, obtains every kind of rarefaction representation corresponding to vitro detection information
Coefficient；
According to described rarefaction representation coefficient, preset training sample and sparse grader, described test sample is examined
Disconnected.
Further, described preset training sample comprises the training sample feature corresponding with every kind of vitro detection sample
Gathering, and described training sample characteristic set is divided into J class training characteristics, J is positive integer and is more than 1；
Described described K kind vitro detection information is carried out rarefaction representation, obtain corresponding sparse of every kind of vitro detection information
Represent that coefficient includes:
According to K the characteristic vector that described K kind vitro detection information architecture is corresponding；
The J class training characteristics corresponding according to described K characteristic vector, every kind of vitro detection sample and preset rarefaction representation
Model carries out rarefaction representation, obtains the rarefaction representation coefficient corresponding with each characteristic vector.
Further, described according to described rarefaction representation coefficient, preset training sample and sparse grader to described survey
Sample originally carries out diagnosis and includes:
According to described K characteristic vector, sparse grader, described training sample and described and described K characteristic vector pair
The rarefaction representation coefficient answered determines the classification belonging to described test sample；
Corresponding diagnostic message is obtained according to described classification.
Further, described K the characteristic vector corresponding according to described K kind vitro detection information architecture includes:
Described K kind vitro detection information is carried out pretreatment respectively；
Extract the characteristic information of pretreated described K kind vitro detection information respectively；
Characteristic information according to the every kind of vitro detection information extracted builds characteristic of correspondence vector.
Further, described method also includes:
When, after the diagnostic message obtaining described user, sending described diagnostic message to user's display interface.
Additionally, for achieving the above object, the present invention also provides for a kind of traditional Chinese medical science in-vitro diagnosis device, described traditional Chinese medical science in-vitro diagnosis
Device includes:
Acquisition module, is used for obtaining test sample, and described test sample is by the K kind vitro detection information of the user got
Composition, K is positive integer and is more than or equal to 2；
Rarefaction representation module, for described K kind vitro detection information is carried out rarefaction representation, obtains every kind of vitro detection letter
The rarefaction representation coefficient that breath is corresponding；
Diagnostic module, is used for according to described rarefaction representation coefficient, preset training sample and sparse grader described survey
Sample originally diagnoses.
Further, described preset training sample comprises the training sample feature corresponding with every kind of vitro detection sample
Gathering, and described training sample characteristic set is divided into J class training characteristics, J is positive integer and is more than 1；
Described rarefaction representation module includes:
Characteristic vector construction unit, for K the characteristic vector corresponding according to described K kind vitro detection information architecture；
Rarefaction representation unit, for special according to described K characteristic vector, J class training that every kind of vitro detection sample is corresponding
Preset sparse representation model of seeking peace carries out rarefaction representation, obtains the rarefaction representation coefficient corresponding with each characteristic vector.
Further, described diagnostic module includes:
Classification confirmation unit, for according to described K characteristic vector, sparse grader, described training sample and described and
Rarefaction representation coefficient corresponding to described K characteristic vector determines the classification belonging to described test sample；
Acquiring unit, for obtaining corresponding diagnostic message according to described classification.
Further, described characteristic vector construction unit includes:
Pretreatment subelement, for carrying out pretreatment respectively to described K kind vitro detection information；
Feature extraction subelement, for extracting the characteristic information of pretreated described K kind vitro detection information respectively；
Characteristic vector builds subelement, builds correspondence for the characteristic information according to the every kind of vitro detection information extracted
Characteristic vector.
Further, described device also includes:
Sending module, for when, after the diagnostic message obtaining described user, sending described diagnostic message and show to user
Interface.
The embodiment of the present invention is by obtaining test sample, and described test sample is by the K kind vitro detection of the user got
Information forms, and K is positive integer and is more than or equal to 2；Described K kind vitro detection information is carried out rarefaction representation, obtain every kind external
The rarefaction representation coefficient that detection information is corresponding；According to described rarefaction representation coefficient, preset training sample and sparse grader pair
Described test sample diagnoses.By obtaining test sample, test sample includes at least two vitro detection information, then by body
Outer detection information carries out rarefaction representation and achieves the fusion of multiple vitro detection information, achieves further according to the classification of sparse grader
Automatically diagnosing according to the vitro detection information after merging, therefore the present invention divides by carrying out human body many-side feature merging
Analysis realizes automatic diagnostic purpose.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of traditional Chinese medical science in-vitro diagnosis method first embodiment of the present invention；
Fig. 2 is the refinement step schematic flow sheet of step S20 in embodiment illustrated in fig. 1 of the present invention；
The refinement step schematic flow sheet of step S30 in Fig. 3 embodiment illustrated in fig. 1 of the present invention；
Fig. 4 is the refinement step schematic flow sheet of step S210 in embodiment illustrated in fig. 2 of the present invention；
Fig. 5 is the high-level schematic functional block diagram of traditional Chinese medical science in-vitro diagnosis device first embodiment of the present invention；
Fig. 6 is the refinement high-level schematic functional block diagram of rarefaction representation module in embodiment illustrated in fig. 5 of the present invention；
Fig. 7 is the refinement high-level schematic functional block diagram of diagnostic module in embodiment illustrated in fig. 5 of the present invention；
Fig. 8 is the refinement high-level schematic functional block diagram of characteristic vector construction unit in embodiment illustrated in fig. 6 of the present invention.
The realization of the object of the invention, functional characteristics and advantage will in conjunction with the embodiments, are described further referring to the drawings.
Detailed description of the invention
Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.
The present invention provides a kind of traditional Chinese medical science in-vitro diagnosis method.With reference to Fig. 1, in the first embodiment, the method includes:
Step S10, obtains test sample, and described test sample is made up of the K kind vitro detection information of the user got,
K is positive integer and is more than or equal to 2；
Step S20, carries out rarefaction representation to described K kind vitro detection information, obtains every kind of vitro detection information corresponding
Rarefaction representation coefficient；
Step S30, according to described rarefaction representation coefficient, preset training sample and sparse grader to described test sample
Diagnose.
The traditional Chinese medical science in-vitro diagnosis method that the present invention provides is for carrying out convergence analysis by the acquisition many features of human body
Realize diagnosis automatically.The traditional Chinese medical science in-vitro diagnosis of the present invention specifically refers to combine detection device based on Chinese traditional medicine principle outside human body and obtains
Take human body many vitro detection information thus judge a kind of diagnostic method of body condition.
Above-mentioned test sample is the test sample for detecting user's fuselage state, and test sample is external by the K kind of user
Detection information forms, and wherein vitro detection information has K kind, and K is positive integer and is more than or equal to 2, shows the vitro detection letter obtained
Breath at least two kinds.Such as vitro detection information can include face image information, lingual surface information, tongue arteries and veins information, pulse condition information, abnormal smells from the patient
At least two information in information and acoustic information.
The method obtaining K kind vitro detection information can be obtained by all kinds of detection devices.Concrete, by taking the photograph
The information such as lingual surface image, tongue arteries and veins image and the face-image as first-class image acquisition device user；Pass through olfactory sensor
The information such as carbon dioxide in perception user's exhaled gas, acetone, ammonia, nitrogen, ethanol；Used by pulse condition collecting device collection
The pulse condition information at family, wherein pulse condition information can be floating, in, any one or more in heavy passage；Pass through voice collection device
Gather the sound of user, when carrying out sound collection, can be acquired in the space closed, to reduce the dry of extraneous factor
Disturb, make collection to sound closer to actual value.
When after the vitro detection information getting user, vitro detection information is carried out rarefaction representation.Rarefaction representation
(sparse representation) is a kind of catabolic process to primary signal, and the purpose that signal carries out rarefaction representation is
The expression way the most succinct in order to obtain signal, thus be easier to obtain the information contained in signal.When carrying out sparse table
By sparse coefficient, K kind vitro detection information is indicated when showing, obtains every kind of rarefaction representation corresponding to vitro detection information
Coefficient, here can be divided into rarefaction representation coefficient two parts, i.e. represent the first sparse coefficient and the representative of similar portion
Second sparse coefficient of characteristic.Similar portion is to represent the dependency between various information, and characteristic is to represent all kinds of letter
Cease respective characteristic.
The method obtaining the sparse coefficient in sparse expression formula according to rarefaction representation has a lot, comes for example with iteration optimization
Solving, concrete operation method can select as required.
After obtaining rarefaction representation coefficient, according to rarefaction representation coefficient, preset training sample and sparse grader pair
Test sample diagnoses.Test sample carries out diagnosis refer to test sample is analyzed, thus diagnostic test sample generation
Which kind of situation of table user's body.
The embodiment of the present invention is by obtaining test sample, and described test sample is by the K kind vitro detection of the user got
Information forms, and K is positive integer and is more than or equal to 2；Described K kind vitro detection information is carried out rarefaction representation, obtain every kind external
The rarefaction representation coefficient that detection information is corresponding；According to described rarefaction representation coefficient, preset training sample and sparse grader pair
Described test sample diagnoses.By obtaining test sample, test sample includes at least two vitro detection information, then by body
Outer detection information carries out rarefaction representation and achieves the fusion of multiple vitro detection information, achieves further according to the classification of sparse grader
Automatically diagnosing according to the vitro detection information after merging, therefore the present invention divides by carrying out human body many-side feature merging
Analysis realizes automatic diagnostic purpose.
Further, described preset training sample comprises the training sample feature corresponding with every kind of vitro detection sample
Gathering, and described training sample characteristic set is divided into J class training characteristics, J is positive integer and is more than 1；With reference to Fig. 2, shown in Fig. 1
The refinement step schematic flow sheet of step S20 in embodiment, above-mentioned steps S20 includes:
Step S210, according to K the characteristic vector that described K kind vitro detection information architecture is corresponding；
Step S220, the J class training characteristics corresponding according to described K characteristic vector, every kind of vitro detection sample and preset
Sparse representation model carry out rarefaction representation, obtain the rarefaction representation coefficient corresponding with each characteristic vector.
The above-mentioned preset training sample of the present embodiment is preset sample characteristics, and training sample comprises and every kind of vitro detection
The sample feature set that sample is corresponding, and described sample feature set is divided into J class, J is positive integer and more than 1, such as, trains sample
Comprising the sample feature set of lingual surface information in Ben, be divided into again J class in the sample feature set of lingual surface information, each class lingual surface is instructed
Practice feature and all represent different body condition.
According to K kind K characteristic vector of vitro detection information architecture, each characteristic vector represents every kind of vitro detection information
Feature.Then according to J class training characteristics corresponding to K characteristic vector, every kind of vitro detection sample and preset rarefaction representation mould
Type carries out rarefaction representation, owing to comprising the training sample feature set corresponding with every kind of vitro detection sample in preset training sample
Close, and training sample is divided into again J class, so every kind of vitro detection sample has corresponding J class training characteristics.Preset is sparse
Represent that model is the model for representing K kind vitro detection information that the principle according to rarefaction representation is arranged.
Concrete, if getting K kind vitro detection information, then the characteristic vector of every kind of vitro detection information is respectively y1、
y2、y3…yk…yK, i.e. test sample y={ y1,y2,y3,…,yk..., yK}, wherein k is the feature of kth kind vitro detection information
Vector.Preset training sample is Dk={ Dk,1,Dk,2,Dk,3,…,Dk,j,…,Dk,J, wherein DkRepresent and kth kind vitro detection
The training sample characteristic set that sample is corresponding, Dk,jRepresent jth class training sample in kth kind training sample characteristic set, total J
Class.If rarefaction representation coefficient is set to represent the first sparse coefficient of similar portionThe second of representative feature part is sparse
CoefficientThen J class training characteristics corresponding to K kind vitro detection information, every kind of vitro detection sample and sparse representation model are entered
Row rarefaction representation is as follows:
WhereinWith For column vector,Represent all categories vitro detection information common portion sparse coefficient
Average, the occurrence of λ with τ regulates according to classification accuracy, from above-mentioned expression it can be seen thatItem ensure that each
There is between individual modal characteristics similar part, remain again its incoherent part simultaneously, by augmentation Lagrange to above-mentioned
The first sparse coefficient in rarefaction representationWith the second sparse coefficientCarry out alternating iteration to solve.
The present embodiment passes through according to K kind vitro detection information architecture characteristic vector, then carries out carrying out sparse to characteristic vector
Represent, thus obtain rarefaction representation coefficient, the multiple vitro detection information got is merged, is to test specimens simultaneously
Originally carry out classification to prepare, after test sample is classified, obtain corresponding diagnostic message, thus realize diagnosis automatically
Purpose.
Further, with reference to Fig. 3, for the refinement step schematic flow sheet of step S30, above-mentioned step in embodiment illustrated in fig. 1
Rapid S30 includes:
Step S310, according to described K characteristic vector, sparse grader, described training sample and described and described K spy
The rarefaction representation coefficient levying vector corresponding determines the classification belonging to described test sample；
Step S320, obtains corresponding diagnostic message according to described classification.
In the present embodiment after getting the rarefaction representation coefficient corresponding with K characteristic vector, according to K characteristic vector,
Sparse grader, training sample and the rarefaction representation coefficient corresponding with K characteristic vector confirm to be made up of K kind vitro detection information
The classification belonging to test sample.Then obtaining corresponding diagnostic message further according to classification, this diagnostic message can be with classification pair
Should be saved in preset memory block.
Specifically K characteristic vector, training sample and the rarefaction representation coefficient corresponding with each characteristic vector are substituted into dilute
Dredge in grader and calculate, confirm when the operation values minimum of sparse grader, corresponding training sample characteristic set class
Classification in not is the classification of described test sample, obtains, further according to described classification, the diagnostic message that the category is corresponding.
Such as, sparse grader can beWherein Dk,jRepresent and kth kind body
Jth class training sample in the kth kind training sample set that outer detection sample is corresponding.WithFor corresponding first sparse coefficient and
Second sparse coefficient.
When classifying, the class selecting K kind reconstructed error value sum minimum is final classification, as j=2, and this expression
Formula is relative to all j！=2 is minimum, i.e.Minimum, then to be made up of K kind vitro detection information
Test sample is just belonging to Equations of The Second Kind, then obtains the diagnostic message that Equations of The Second Kind is corresponding.The class choosing minima is because, should
What sparse grader represented is the similarity of test sample training sample, when difference is minimum, shows in this test sample external in K
Detection information is the most similar to training sample, therefore selects the diagnostic message that the most similar training sample is corresponding as this user
Diagnostic message.
The present embodiment is by according to K characteristic vector, sparse grader and training sample and corresponding with K characteristic vector
Rarefaction representation coefficient, determines the classification of test sample, it is achieved that classify the test sample of the fuselage state representing user,
Thus obtain corresponding diagnostic message according to classification, it is achieved that obtain the automation process of diagnostic result.
Further, with reference to Fig. 4, for the refinement step schematic flow sheet of step S210, above-mentioned step in embodiment illustrated in fig. 2
Rapid S210 includes:
Step S211, carries out pretreatment respectively to described K kind vitro detection information；
Step S212, extracts the characteristic information of pretreated described K kind vitro detection information respectively；
Step S213, builds characteristic of correspondence vector according to the characteristic information of the every kind of vitro detection information extracted.
In the present embodiment after getting K kind vitro detection information, K kind vitro detection information is carried out pretreatment.Pre-place
Reason process is to prepare characteristic information extraction.Such as, if the vitro detection information got comprises image, then to image
Carry out color correction；If the vitro detection information got comprises pulse condition information, then judge the dimension of pulse condition information, if dimension mistake
Height, then carry out dimension-reduction treatment, PCA wherein can be used to carry out dimensionality reduction pulse condition information；If the external inspection got
Measurement information comprises acoustic information, then acoustic information is entered denoising；If the vitro detection information got comprises olfactory information,
Then olfactory information is carried out denoising etc..The concrete method taking pretreatment can be determined as required.
After K kind vitro detection information is carried out pretreatment, extract the spy of pretreated K kind vitro detection information respectively
Reference ceases, and i.e. every kind of vitro detection information is carried out feature extraction, and the method for concrete feature extraction can be carried out as required
Select, as geometric method, modelling, signal processing method used to carry out feature extraction, for different classes of for image information
Vitro detection information different methods can be used to carry out feature extraction.
When after the characteristic information being extracted every kind of vitro detection information, according to the spy of the every kind of vitro detection information extracted
Levying information architecture characteristic of correspondence vector, represent in the most each vector is the characteristic information of this kind of vitro detection information.
The present embodiment by first carrying out pretreatment, then characteristic information extraction to K kind vitro detection information, and then builds K
Characteristic vector so that the characteristic vector of structure can represent K kind vitro detection information the most accurately, improves the accurate of diagnosis
Degree.
Further, the method that the present invention proposes also includes:
When, after the diagnostic message obtaining described user, sending described diagnostic message to user's display interface.
When, after the diagnostic message being obtained user by said method, sending diagnostic message and show to user in the present embodiment
Interface, so that user's display interface shows this diagnostic message.Concrete, in conjunction with display device diagnostic message can be shown and examine
In disconnected information, this diagnostic message can show by preset form, it is also possible to for diagnostic message client pointed out or to
Give relevant medical science suggestion etc..
It is understood that except diagnostic message can be sent to user's display interface, so that user's display interface shows
This diagnostic message, it is also possible to send diagnostic result to user by modes such as mails, or remind user by alerting pattern
Check diagnostic result.
After the present embodiment is by obtaining the diagnostic message of user, send diagnostic message to user's display interface so that use
Diagnostic message can be checked in family, it is achieved that from the process obtaining the whole automatization of result being detected, makes diagnosis process the most convenient,
Improve Consumer's Experience.
The present invention also provides for a kind of traditional Chinese medical science in-vitro diagnosis device, with reference to Fig. 5, it is provided that traditional Chinese medical science in-vitro diagnosis device of the present invention
First embodiment, in this embodiment, traditional Chinese medical science in-vitro diagnosis device includes:
Acquisition module 10, is used for obtaining test sample, and described test sample is believed by the K kind vitro detection of the user got
Breath composition, K is positive integer and is more than or equal to 2；
Rarefaction representation module 20, for described K kind vitro detection information is carried out rarefaction representation, obtains every kind of vitro detection
The rarefaction representation coefficient that information is corresponding；
Diagnostic module 30, is used for according to described rarefaction representation coefficient, preset training sample and sparse grader described
Test sample diagnoses.
The traditional Chinese medical science in-vitro diagnosis device that the present invention provides is for carrying out convergence analysis by the acquisition many features of human body
Realize diagnosis automatically.The traditional Chinese medical science in-vitro diagnosis of the present invention specifically refers to combine detection device based on Chinese traditional medicine principle outside human body and obtains
Take human body many vitro detection information thus judge a kind of diagnostic method of body condition.
Above-mentioned test sample is the test sample for detecting user's fuselage state, and test sample is external by the K kind of user
Detection information forms, and wherein vitro detection information has K kind, and K is positive integer and is more than or equal to 2, shows the vitro detection letter obtained
Breath at least two kinds.Such as vitro detection information can include face image information, lingual surface information, tongue arteries and veins information, pulse condition information, abnormal smells from the patient
At least two information in information and acoustic information.
Acquisition module 10 is obtained the method for K kind vitro detection information and can be obtained by all kinds of detection devices.Tool
Body, by information such as the shooting lingual surface image of first-class image acquisition device user, tongue arteries and veins image and face-images；Logical
Cross the information such as the carbon dioxide in olfactory sensor perception user's exhaled gas, acetone, ammonia, nitrogen, ethanol；Adopted by pulse condition
Acquisition means gathers the pulse condition information of user, wherein pulse condition information can be floating, in, sink any one or more in passage；Pass through
Voice collection device gathers the sound of user, when carrying out sound collection, can be acquired in the space closed, to reduce
The interference of extraneous factor, make collection to sound closer to actual value.
When, after the vitro detection information getting user, rarefaction representation module 20 carries out sparse table to vitro detection information
Show.Rarefaction representation (sparse representation) is a kind of catabolic process to primary signal, and signal is carried out sparse table
The purpose shown is the expression way the most succinct in order to obtain signal, thus is easier to obtain the information contained in signal.When
By sparse coefficient, K kind vitro detection information is indicated when carrying out rarefaction representation, obtains every kind of vitro detection information corresponding
Rarefaction representation coefficient, rarefaction representation coefficient here can be divided into two parts, i.e. represent the first sparse of similar portion
Coefficient and the second sparse coefficient of representative feature part.Similar portion is to represent the dependency between various information, and characteristic is
Represent the respective characteristic of various information.
The method obtaining the sparse coefficient in sparse expression formula according to rarefaction representation has a lot, comes for example with iteration optimization
Solving, concrete operation method can select as required.
After obtaining rarefaction representation coefficient, diagnostic module 30 is according to rarefaction representation coefficient, preset training sample and dilute
Dredge grader test sample is diagnosed.Test sample carries out diagnosis refer to test sample is analyzed, thus diagnose
Test sample represents which kind of situation of user's body.
The embodiment of the present invention is by obtaining test sample, and described test sample is by the K kind vitro detection of the user got
Information forms, and K is positive integer and is more than or equal to 2；Described K kind vitro detection information is carried out rarefaction representation, obtain every kind external
The rarefaction representation coefficient that detection information is corresponding；According to described rarefaction representation coefficient, preset training sample and sparse grader pair
Described test sample diagnoses.By obtaining test sample, test sample includes at least two vitro detection information, then by body
Outer detection information carries out rarefaction representation and achieves the fusion of multiple vitro detection information, achieves further according to the classification of sparse grader
Automatically diagnosing according to the vitro detection information after merging, therefore the present invention divides by carrying out human body many-side feature merging
Analysis realizes automatic diagnostic purpose.
Further, described preset training sample comprises the training sample feature corresponding with every kind of vitro detection sample
Gathering, and described training sample characteristic set is divided into J class training characteristics, J is positive integer and is more than 1；
With reference to Fig. 6, for the refinement high-level schematic functional block diagram of rarefaction representation module described in embodiment illustrated in fig. 5 20, described
Rarefaction representation module 20 includes:
Characteristic vector construction unit 210, for K the characteristic vector corresponding according to described K kind vitro detection information architecture；
Rarefaction representation unit 220, for training according to described K characteristic vector, J class that every kind of vitro detection sample is corresponding
Feature and preset sparse representation model carry out rarefaction representation, obtain the rarefaction representation coefficient corresponding with each characteristic vector.
The above-mentioned preset training sample of the present embodiment is preset sample characteristics, and training sample comprises and every kind of vitro detection
The sample feature set that sample is corresponding, and described sample feature set is divided into J class, J is positive integer and more than 1, such as, trains sample
Comprising the sample feature set of lingual surface information in Ben, be divided into again J class in the sample feature set of lingual surface information, each class lingual surface is instructed
Practice feature and all represent different body condition.
K characteristic vector, Mei Gete is built according to characteristic vector construction unit 210 after getting K kind vitro detection information
Levy vector and represent the feature of every kind of vitro detection information.Rarefaction representation unit 220 is according to K characteristic vector, every kind of vitro detection
J class training characteristics and preset sparse representation model that sample is corresponding carry out rarefaction representation, obtain corresponding with each characteristic vector
Rarefaction representation coefficient, owing to preset training sample comprising the training sample feature set corresponding with every kind of vitro detection sample
Close, and training sample is divided into again J class, so every kind of vitro detection sample has corresponding J class training characteristics.Preset is sparse
Represent that model is the model for representing K kind vitro detection information that the principle according to rarefaction representation is arranged.
Concrete, if getting K kind vitro detection information, then the characteristic vector of every kind of vitro detection information is respectively y1、
y2、y3…yk…yK, i.e. test sample y={ y1,y2,y3,…,yk,…,yK, wherein k is the feature of kth kind vitro detection information
Vector.Preset training sample is Dk={ Dk,1,Dk,2,Dk,3,…,Dk,j,…,Dk,J, wherein DkRepresent and kth kind vitro detection
The training sample characteristic set that sample is corresponding, Dk,jRepresent jth class training sample in kth kind training sample characteristic set, total J
Class.If rarefaction representation coefficient is set to represent the first sparse coefficient of similar portionThe second of representative feature part is sparse
CoefficientThen J class training characteristics corresponding to K kind vitro detection information, every kind of vitro detection sample and sparse representation model are entered
Row rarefaction representation is as follows:
WhereinWith For column vector,Represent all categories vitro detection information common portion sparse coefficient
Average, the occurrence of λ with τ regulates according to classification accuracy, from above-mentioned expression it can be seen thatItem ensure that each
There is between individual modal characteristics similar part, remain again its incoherent part simultaneously, by augmentation Lagrange to above-mentioned
The first sparse coefficient in rarefaction representationWith the second sparse coefficientCarry out alternating iteration to solve.
The present embodiment passes through according to K kind vitro detection information architecture characteristic vector, then carries out carrying out sparse to characteristic vector
Represent, thus obtain rarefaction representation coefficient, the multiple vitro detection information got is merged, is to test specimens simultaneously
Originally carry out classification to prepare, after test sample is classified, obtain corresponding diagnostic message, thus realize diagnosis automatically
Purpose.
Further, with reference to Fig. 7, for the refinement high-level schematic functional block diagram of diagnostic module 30 described in embodiment illustrated in fig. 5,
Described diagnostic module 30 includes:
Classification confirmation unit 310, for according to described K characteristic vector, sparse grader, described training sample and described
The rarefaction representation coefficient corresponding with described K characteristic vector determines the classification belonging to described test sample；
Acquiring unit 320, for obtaining corresponding diagnostic message according to described classification.
In the present embodiment after getting the rarefaction representation coefficient corresponding with K characteristic vector, classification confirmation unit 310
Confirm by K kind body according to K characteristic vector, sparse grader, training sample and the rarefaction representation coefficient corresponding with K characteristic vector
The classification belonging to test sample of outer detection information composition, specifically determines that the fuselage state that K kind vitro detection information represents belongs to
In which classification.Second acquisition unit 320 obtains corresponding diagnostic message according to classification, and this diagnostic message can be corresponding with classification
It is saved in preset memory block.
Specifically K characteristic vector, training sample and the rarefaction representation coefficient corresponding with each characteristic vector are substituted into dilute
Dredge in grader and calculate, confirm when the operation values minimum of sparse grader, corresponding training sample characteristic set class
Classification in not is the classification of described test sample, obtains, further according to described classification, the diagnostic message that the category is corresponding.
Such as, sparse grader can beWherein Dk,jRepresent and kth kind body
Jth class training sample in the kth kind training sample set that outer detection sample is corresponding.WithFor corresponding first sparse coefficient and
Second sparse coefficient.
When classifying, the class selecting K kind reconstructed error value sum minimum is final classification, as j=2, and this expression
Formula is relative to all j！=2 is minimum, i.e.Minimum, then to be made up of K kind vitro detection information
Test sample is just belonging to Equations of The Second Kind, then obtains the diagnostic message that Equations of The Second Kind is corresponding.The class choosing minima is because, should
What sparse grader represented is the similarity of test sample training sample, when difference is minimum, shows in this test sample external in K
Detection information is the most similar to training sample, therefore selects the diagnostic message that the most similar training sample is corresponding as this user
Diagnostic message.
The present embodiment is by according to K characteristic vector, sparse grader and training sample and corresponding with K characteristic vector
Rarefaction representation coefficient, determines the classification of test sample, it is achieved that classify the test sample of the fuselage state representing user,
Thus obtain corresponding diagnostic message according to classification, it is achieved that obtain the automation process of diagnostic result.
Further, with reference to Fig. 8, the refinement functional module signal of characteristic vector construction unit 210 in embodiment illustrated in fig. 6
Figure, described characteristic vector construction unit 210 includes:
Pretreatment subelement 211, for carrying out pretreatment respectively to described K kind vitro detection information；
Feature extraction subelement 212, for extracting the feature letter of pretreated described K kind vitro detection information respectively
Breath；
Characteristic vector builds subelement 213, builds for the characteristic information according to the every kind of vitro detection information extracted
Characteristic of correspondence vector.
In the present embodiment after getting K kind vitro detection information, pretreatment subelement 211 is to K kind vitro detection information
Carry out pretreatment.Preprocessing process is to prepare characteristic information extraction.Such as, if the vitro detection information bag got
Containing image, then image is carried out color correction；If the vitro detection information got comprises pulse condition information, then judge pulse condition information
Dimension, if dimension is too high, then pulse condition information is carried out dimension-reduction treatment, PCA wherein can be used to carry out dimensionality reduction；
If the vitro detection information got comprises acoustic information, then acoustic information is entered denoising；If the vitro detection got
Information comprises olfactory information, then olfactory information is carried out denoising etc..The concrete method taking pretreatment can be according to need
It is determined.
After K kind vitro detection information is carried out pretreatment, feature extraction subelement 212 extracts pretreated K respectively
Plant the characteristic information of vitro detection information, i.e. every kind of vitro detection information is carried out feature extraction, the side of concrete feature extraction
Method can select as required, as geometric method, modelling, signal processing method used to carry out feature for image information
Extract, different methods can be used to carry out feature extraction for different classes of vitro detection information.
When, after the characteristic information being extracted every kind of vitro detection information, characteristic vector builds subelement 213 according to extracting
Every kind of vitro detection information characteristic information build characteristic of correspondence vector, in the most each vector represent be this kind of external inspection
The characteristic information of measurement information.
The present embodiment by first carrying out pretreatment, then characteristic information extraction to K kind vitro detection information, and then builds K
Characteristic vector so that the characteristic vector of structure can represent K kind vitro detection information the most accurately, improves the accurate of diagnosis
Degree.
Further, in the present embodiment, described device also includes:
Sending module, for when, after the diagnostic message obtaining described user, sending described diagnostic message and show to user
Interface.
When, after the diagnostic message being obtained user by said method, sending module sends diagnostic message extremely in the present embodiment
User's display interface, so that user's display interface shows this diagnostic message.Concrete, can be in conjunction with display device by diagnostic message
Display is in diagnostic message, and this diagnostic message can show by preset form, it is also possible to carries client for diagnostic message
The medical science suggestion etc. shown or be correlated with.
It is understood that except diagnostic message can be sent to user's display interface, so that user's display interface shows
This diagnostic message, it is also possible to send diagnostic result to user by modes such as mails, or remind user by alerting pattern
Check diagnostic result.
After the present embodiment is by obtaining the diagnostic message of user, send diagnostic message to user's display interface so that use
Diagnostic message can be checked in family, it is achieved that from the process obtaining the whole automatization of result being detected, makes diagnosis process the most convenient,
Improve Consumer's Experience.
These are only the preferred embodiments of the present invention, not thereby limit the scope of the claims of the present invention, every utilize this
Equivalent structure or equivalence flow process that bright description and accompanying drawing content are made convert, or are directly or indirectly used in other relevant skills
Art field, is the most in like manner included in the scope of patent protection of the present invention.
Claims (10)
1. a traditional Chinese medical science in-vitro diagnosis method, it is characterised in that said method comprising the steps of:
Obtain test sample, described test sample is made up of the K kind vitro detection information of the user got, K be positive integer and
More than or equal to 2；
Described K kind vitro detection information is carried out rarefaction representation, obtains every kind of rarefaction representation coefficient corresponding to vitro detection information；
According to described rarefaction representation coefficient, preset training sample and sparse grader, described test sample is diagnosed.
2. the method for claim 1, it is characterised in that comprise and every kind of vitro detection in described preset training sample
The training sample characteristic set that sample is corresponding, and described training sample characteristic set is divided into J class training characteristics, J be positive integer and
More than 1；
Described described K kind vitro detection information is carried out rarefaction representation, obtain every kind of rarefaction representation corresponding to vitro detection information
Coefficient includes:
According to K the characteristic vector that described K kind vitro detection information architecture is corresponding；
The J class training characteristics corresponding according to described K characteristic vector, every kind of vitro detection sample and preset sparse representation model
Carry out rarefaction representation, obtain the rarefaction representation coefficient corresponding with each characteristic vector.
3. method as claimed in claim 2, it is characterised in that described according to described rarefaction representation coefficient, preset training sample
This and sparse grader carry out diagnosis to described test sample and include:
Corresponding with described and described K characteristic vector according to described K characteristic vector, sparse grader, described training sample
Rarefaction representation coefficient determines the classification belonging to described test sample；
Corresponding diagnostic message is obtained according to described classification.
4. method as claimed in claim 2, it is characterised in that the described K corresponding according to described K kind vitro detection information architecture
Individual characteristic vector includes:
Described K kind vitro detection information is carried out pretreatment respectively；
Extract the characteristic information of pretreated described K kind vitro detection information respectively；
Characteristic information according to the every kind of vitro detection information extracted builds characteristic of correspondence vector.
5. the method as described in any one of claim 1-4, it is characterised in that described method also includes:
When, after the diagnostic message obtaining described user, sending described diagnostic message to user's display interface.
6. a traditional Chinese medical science in-vitro diagnosis device, it is characterised in that described device includes:
Acquisition module, is used for obtaining test sample, and described test sample is by the K kind vitro detection information group of the user got
Becoming, K is positive integer and is more than or equal to 2；
Rarefaction representation module, for described K kind vitro detection information is carried out rarefaction representation, obtains every kind of vitro detection information pair
The rarefaction representation coefficient answered；
Diagnostic module, is used for according to described rarefaction representation coefficient, preset training sample and sparse grader described test specimens
Originally diagnose.
7. device as claimed in claim 6, it is characterised in that comprise and every kind of vitro detection in described preset training sample
The training sample characteristic set that sample is corresponding, and described training sample characteristic set is divided into J class training characteristics, J be positive integer and
More than 1；
Described rarefaction representation module includes:
Characteristic vector construction unit, for K the characteristic vector corresponding according to described K kind vitro detection information architecture；
Rarefaction representation unit, for the J class training characteristics corresponding according to described K characteristic vector, every kind of vitro detection sample and
Preset sparse representation model carries out rarefaction representation, obtains the rarefaction representation coefficient corresponding with each characteristic vector.
8. device as claimed in claim 7, it is characterised in that described diagnostic module includes:
Classification confirmation unit, for according to described K characteristic vector, sparse grader, described training sample and described and described K
Rarefaction representation coefficient corresponding to individual characteristic vector determines the classification belonging to described test sample；
Acquiring unit, for obtaining corresponding diagnostic message according to described classification.
9. device as claimed in claim 7, it is characterised in that described characteristic vector construction unit includes:
Pretreatment subelement, for carrying out pretreatment respectively to described K kind vitro detection information；
Feature extraction subelement, for extracting the characteristic information of pretreated described K kind vitro detection information respectively；
Characteristic vector builds subelement, builds corresponding spy for the characteristic information according to the every kind of vitro detection information extracted
Levy vector.
10. the device as described in any one of claim 7 to 9, it is characterised in that described device also includes:
Sending module, for when, after the diagnostic message obtaining described user, sending described diagnostic message to user's display interface.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610377994.8A CN106055896B (en) | 2016-05-31 | 2016-05-31 | Chinese medicine in-vitro diagnosis method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610377994.8A CN106055896B (en) | 2016-05-31 | 2016-05-31 | Chinese medicine in-vitro diagnosis method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106055896A true CN106055896A (en) | 2016-10-26 |
| CN106055896B CN106055896B (en) | 2019-04-05 |
Family
ID=57171688
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610377994.8A Expired - Fee Related CN106055896B (en) | 2016-05-31 | 2016-05-31 | Chinese medicine in-vitro diagnosis method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106055896B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223700A (en) * | 2021-04-14 | 2021-08-06 | 北京联世科技有限公司 | Traditional Chinese medicine pulse condition identification method and device based on pulse condition data |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1759804A (en) * | 2005-11-02 | 2006-04-19 | 浙江大学 | Intelligent analyzing and differentiating method of herbalist doctor through integrated references form four parts of diagnosis |
| CN101576913B (en) * | 2009-06-12 | 2011-09-21 | 中国科学技术大学 | Automatic clustering, visual and retrieval system for tongue picture based on self-organizing map neural network |
-
2016
- 2016-05-31 CN CN201610377994.8A patent/CN106055896B/en not_active Expired - Fee Related
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223700A (en) * | 2021-04-14 | 2021-08-06 | 北京联世科技有限公司 | Traditional Chinese medicine pulse condition identification method and device based on pulse condition data |
| CN113223700B (en) * | 2021-04-14 | 2022-02-08 | 北京联世科技有限公司 | Traditional Chinese medicine pulse condition identification method and device based on pulse condition data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106055896B (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106725532B (en) | Depression automatic evaluation system and method based on phonetic feature and machine learning | |
| TWI684997B (en) | Establishing method of bone age assessment and height prediction model, bone age assessment and height prediction system, and bone age assessment and height prediction method | |
| CN101667289B (en) | Retinal image segmentation method based on NSCT feature extraction and supervised classification | |
| CN109480864A (en) | A kind of schizophrenia automatic evaluation system based on nervous functional defects and machine learning | |
| CN108511055A (en) | Ventricular premature beat identifying system and method based on Multiple Classifier Fusion and diagnostic rule | |
| KR102271558B1 (en) | Artificial intelligence scalp image diagnostic analysis system using big data and product recommendation system using the same | |
| CN110731773A (en) | abnormal electrocardiogram screening method based on fusion of global and local depth features of electrocardiogram | |
| Hasan et al. | RGB pixel analysis of fingertip video image captured from sickle cell patient with low and high level of hemoglobin | |
| CN113903082A (en) | Human body gait monitoring algorithm based on dynamic time planning | |
| Çallı et al. | FRODO: Free rejection of out-of-distribution samples: application to chest x-ray analysis | |
| CN110415818A (en) | A kind of intelligent pediatric disease interrogation system and method based on observable illness | |
| CN114358194A (en) | Gesture tracking based detection method for abnormal limb behaviors of autism spectrum disorder | |
| Tan et al. | Automated detection of congenital heart disease in fetal ultrasound screening | |
| CN108511064A (en) | The system for automatically analyzing healthy data based on deep learning | |
| CN110693510A (en) | Attention deficit hyperactivity disorder auxiliary diagnosis device and using method thereof | |
| CN111413492A (en) | Method and system for detecting novel coronavirus COVID-2019 pneumonia | |
| CN107970027A (en) | A kind of radial artery detection and human body constitution identifying system and method | |
| CN113128585B (en) | Deep neural network based multi-size convolution kernel method for realizing electrocardiographic abnormality detection and classification | |
| Kadhim et al. | Detection of COVID-19 in X-Rays by Convolutional Neural Networks | |
| CN106778830A (en) | Staging method based on double cluster results and AdaBoost | |
| CN109223006A (en) | A kind of schizophrenia diagnosis system | |
| CN106055896A (en) | TCM (Traditional Chinese Medicine) in-vitro diagnosis method and device | |
| KR102033484B1 (en) | Method and apparatus for setting normal reference range in clinical inspection of pets using generative adversary network | |
| CN116740426A (en) | Classification prediction system for functional magnetic resonance images | |
| CN110265140A (en) | Foot deformity detection model, foot deformity detection system and foot deformity detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 518000 Guangdong city of Shenzhen province Nanshan District Taoyuan Xili Street University City Building 3 layer 4 Valley two Applicant after: Shenzhen Zhong Shi Health Science and Technology Co., Ltd. Address before: 510000 Guangdong city of Shenzhen province Qianhai Shenzhen Hong Kong cooperation zone before Bay Road No. 1 building 201 room A (located in Shenzhen Qianhai business secretary Co. Ltd.) Applicant before: Shenzhen Creative Technology Limited |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20181122 Address after: 518000 Room 201, building A, No. 1, Qian Wan Road, Qianhai Shenzhen Hong Kong cooperation zone, Shenzhen, Guangdong (Shenzhen Qianhai business secretary Co., Ltd.) Applicant after: Shenzhen Lizhong Mdt InfoTech Ltd Address before: 518000 Guangdong Shenzhen Nanshan District Taoyuan street, Xili University Town, cloud Valley two phase 4 4 Story 3 floors. Applicant before: Shenzhen Zhong Shi Health Science and Technology Co., Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190405 Termination date: 20190531 |