CN104951666A - Disease diagnosis method and device - Google Patents

Abstract

The invention discloses a disease diagnosis method and device. According to the method, a congeneric similar sample set and a heterogeneous similar sample set of any one training sample are determined on the basis of the cosine similarity principle and according to gene expression data of multiple training samples. Further, a projection matrix of a low-dimensional characteristic space is constructed according to the congeneric similar sample set and the heterogeneous similar sample set and dimensionality reduction processing of the training sample and a test sample is realized by using the projection matrix, so that the test sample can be diagnosed conveniently. Compared with the prior art, the method has the advantages that similarity among samples is measured by using cosine, the measurement precision of the similarity is higher as compared with that of an Euclidean distance manner, so that the disease diagnosis precision is improved.

Description

A kind of methods for the diagnosis of diseases and device

Technical field

The application relates to medical domain, more particularly, relates to a kind of methods for the diagnosis of diseases and device.

Background technology

Along with the development of science and technology, the diagnosis of disease can the help of computer, adopts the diagnosis that the method for machine learning realizes disease.

The data of medical diagnosis on disease are generally the gene expression data of human body, and these data are typical high dimensional data, namely comprise multiple feature.In order to reduce computation complexity, storage complexity, the Dimensionality Reduction of gene expression data is absolutely necessary step.Current main employing is based on the differentiation neighbour embedding grammar of digraph pattern, and the method can carry out dimensionality reduction to data effectively.The method adopts the similar similar sample set of Euclidean distance determination training sample sample set similar with foreign peoples when determining projection matrix, its diagnostic accuracy is low.

Summary of the invention

In view of this, the invention provides a kind of methods for the diagnosis of diseases and device, to improve the diagnostic accuracy of disease.

For achieving the above object, the invention provides following technical scheme:

A kind of methods for the diagnosis of diseases, comprising:

Obtain the gene expression data of test sample book and multiple training sample;

Based on cosine similarity principle, determine the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample;

According to described similar similar sample set sample set similar with described foreign peoples, set up the projection matrix of low dimensional feature space according to presetting method;

Utilize described projection matrix that multiple training sample after test sample book after process and process is mapped to low dimensional feature space;

In low dimensional feature space, determine the training sample nearest with test sample book, give test sample book, to complete the diagnosis to test sample book by the classification of this training sample.

Preferably, described based on cosine similarity principle, determine the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample, also comprise before:

Random parameters extraction process and characteristic value normalization process are carried out to the gene expression data of test sample book and multiple training sample.

Preferably, described according to described similar similar sample set sample set similar with described foreign peoples, set up the projection matrix of low dimensional feature space according to presetting method, comprising:

According to similar similar sample sample similar with foreign peoples, similar matrix between similar matrix and class in structure class;

Wherein:

similar matrix in representation class, similar matrix between representation class, x ' _ithe gene expression data of the training sample i after process, x ' _jrepresent the gene expression data of the rear training sample j of process, represent the similar similar sample set of training sample j, the similar similar sample set of training sample i, represent the similar sample set of foreign peoples of training sample j, represent the similar sample set of foreign peoples of training sample i;

According to similar matrix between similar matrix and class in class, to build between class local Scatter Matrix in local Scatter Matrix and class;

Wherein, S _w=X (D ^w-F ^w) X ^t, S _b=X (D ^b-F ^b) X ^t;

S _wlocal Scatter Matrix in representation class, S _blocal Scatter Matrix between representation class, X represents the gene expression data of sample, F ^wsimilar matrix in representation class, F ^bsimilar matrix between representation class, D ^wand D ^brepresent diagonal matrix,

Feature decomposition is carried out to local Scatter Matrix in local Scatter Matrix between described class and described class, ensures that the ratio of between class distance and inter-object distance maximizes, obtain several eigenwerts according to order from big to small;

The projection matrix of low dimensional feature space is set up according to eigenwert characteristic of correspondence vector.

A kind of medical diagnosis on disease device, comprising:

Data acquisition unit, for obtaining the gene expression data of test sample book and multiple training sample;

Similar Sample Establishing unit similar with foreign peoples, for based on cosine similarity principle, determines the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample;

Projection matrix sets up unit, for according to described similar similar sample set sample set similar with described foreign peoples, sets up the projection matrix of low dimensional feature space according to presetting method;

Map unit, is mapped to low dimensional feature space for utilizing described projection matrix by test sample book and multiple training sample;

Diagnosis unit, in low dimensional feature space, determines the training sample nearest with test sample book, gives test sample book, to complete the diagnosis to test sample book by the classification of this training sample.

Preferably, also comprise: data processing unit, extract process and characteristic value normalization process for carrying out random parameters to the gene expression data of test sample book and multiple training sample.

Preferably, described projection matrix is set up unit and is comprised:

Similar matrix construction unit between similar matrix and class in class, for according to similar similar sample sample similar with foreign peoples, to build in class similar matrix between similar matrix and class;

Wherein:

similar matrix in representation class, similar matrix between representation class, x ' _ithe gene expression data of the training sample i after process, x ' _jthe gene expression data of training sample j is represented after process, represent the similar similar sample set of training sample j, the similar similar sample set of training sample i, represent the similar sample set of foreign peoples of training sample j, represent the similar sample set of foreign peoples of training sample i;

To build between class local Scatter Matrix construction unit in local Scatter Matrix and class, for according to similar matrix between similar matrix and class in class, to build between class local Scatter Matrix in local Scatter Matrix and class;

Wherein, S _w=X (D ^w-F ^w) X ^t, S _b=X (D ^b-F ^b) X ^t;

S _wlocal Scatter Matrix in representation class, S _blocal Scatter Matrix between representation class, X represents the gene expression data of sample, F ^wsimilar matrix in representation class, F ^bsimilar matrix between representation class, D ^wand D ^brepresent diagonal matrix,

Feature decomposition unit, for carrying out feature decomposition to local Scatter Matrix in local Scatter Matrix between described class and described class, ensureing that the ratio of between class distance and inter-object distance maximizes, obtaining several eigenwerts according to order from big to small;

Projection matrix sets up subelement, for setting up the projection matrix of low dimensional feature space according to eigenwert characteristic of correspondence vector.

Known via above-mentioned technical scheme, compared with prior art, the invention discloses a kind of methods for the diagnosis of diseases and device.The method, based on cosine similarity principle, determines the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample.And then, build the projection matrix of low dimensional feature space according to similar similar sample set sample set similar with foreign peoples, utilize the dimension-reduction treatment that this projection matrix realizes training sample and test sample book, to facilitate the diagnosis to test sample book.Compared with prior art, the present invention measures the similarity between sample at employing cosine, and compared with adopting the mode of Euclidean distance, the accuracy of measurement of its similarity is higher, thus improves the precision of medical diagnosis on disease.

Accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present application or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only the embodiment of the application, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to the accompanying drawing provided.

Fig. 1 shows the schematic flow sheet of a kind of methods for the diagnosis of diseases disclosed in one embodiment of the invention;

Fig. 2 shows the structural representation of a kind of medical diagnosis on disease device disclosed in another embodiment of the present invention.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present application, be clearly and completely described the technical scheme in the embodiment of the present application, obviously, described embodiment is only some embodiments of the present application, instead of whole embodiments.Based on the embodiment in the application, those of ordinary skill in the art are not making the every other embodiment obtained under creative work prerequisite, all belong to the scope of the application's protection.

The schematic flow sheet of a kind of methods for the diagnosis of diseases disclosed in one embodiment of the invention is shown see Fig. 1.

As shown in Figure 1, the method comprises:

101: the gene expression data obtaining test sample book and multiple training sample.

If existing gene expression training data is wherein x _i∈ R ^dthe gene expression data of i-th people, y _i={+1 ,-1} represents x _iclass label, N represents number of samples, and D represents the dimension of training data.

102: based on cosine similarity principle, determine the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample.

Need to say, due to for number of samples, the dimension of training data makes a very large number.Thus in other embodiments of the invention in order to reduce computation complexity, before determining similar similar sample set sample set similar with foreign peoples, needing that random character is carried out to training sample and test sample book extract process and characteristic value normalization process.

Such as, extract d feature in training sample, record extracts the position of feature and | I|=d.Then the eigenwert extracted is normalized, makes the scope of eigenwert interval in [0,1].Then remember that random character is selected and training dataset after normalization is and x _i' ∈ R ^d.

And then, according to the feature locations collection I retained and the normalized mode of training sample, construct a new test sample book.

103: according to described similar similar sample set sample set similar with described foreign peoples, set up the projection matrix of low dimensional feature space according to presetting method.

This process specifically comprises the following steps:

A: according to similar similar sample sample similar with foreign peoples, similar matrix between similar matrix and class in structure class,

similar matrix in representation class, similar matrix between representation class, x ' _ithe gene expression data of the training sample i after process, x ' _jrepresent the gene expression data of the rear training sample j of process, represent the similar similar sample set of training sample j, the similar similar sample set of training sample i, represent the similar sample set of foreign peoples of training sample j, represent the similar sample set of foreign peoples of training sample i;

B: according to similar matrix between similar matrix and class in class, to build between class local Scatter Matrix in local Scatter Matrix and class.

Wherein, S _w=X (D ^w-F ^w) X ^t, S _b=X (D ^b-F ^b) X ^t;

S _wlocal Scatter Matrix between representation class, S _blocal Scatter Matrix between representation class, X represents the gene expression data of sample, F ^wsimilar matrix in representation class, F ^bsimilar matrix between representation class, D ^wand D ^brepresent diagonal matrix,

C: feature decomposition is carried out to local Scatter Matrix in local Scatter Matrix between described class and described class, ensures that the ratio of between class distance and inter-object distance maximizes, obtain several eigenwerts according to order from big to small.

D: the projection matrix setting up low dimensional feature space according to eigenwert characteristic of correspondence vector.

In order to obtain projection matrix P, we are to S _band S _wcarry out generalized eigen decomposition.The eigenwert obtained is sorted according to order from big to small, its r eigenwert characteristic of correspondence vector composition matrix P=[p before getting ₁, p ₂, L, p _r], wherein p _iit is the proper vector after feature decomposition.Wherein, r " d " D.

104: utilize described projection matrix that test sample book and multiple training sample are mapped to low dimensional feature space.

After obtaining projection matrix P, the training sample of former sample space is projected to low dimensional feature space, z by projection _i=P ^tx _i', wherein z _ix _i' in the projection of lower dimensional space, z _i∈ R ^r.Order for the training sample after projection.

In like manner, test sample book is projected in low dimensional feature space by projection matrix, obtain the test sample book after projecting.

105: in low dimensional feature space, determine the training sample nearest with test sample book, give test sample book, to complete the diagnosis to test sample book by the classification of this training sample.

Utilize nearest neighbor classifier, the test sample book after projection is differentiated in low dimensional feature space.That is, at training sample in, find the sample nearest with test sample book, and then give projective tests sample the classification of this sample.So just complete the diagnosis to test sample book.

As seen from the above embodiment: the method, based on cosine similarity principle, determines the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample.And then, build the projection matrix of low dimensional feature space according to similar similar sample set sample set similar with foreign peoples, utilize the dimension-reduction treatment that this projection matrix realizes training sample and test sample book, to facilitate the diagnosis to test sample book.Compared with prior art, the present invention measures the similarity between sample at employing cosine, and compared with adopting the mode of Euclidean distance, the accuracy of measurement of its similarity is higher, thus improves the precision of medical diagnosis on disease.

The structural representation of a kind of medical diagnosis on disease device disclosed in another embodiment of the present invention is shown see Fig. 2.

As shown in Figure 2, this device comprises: data acquisition unit 1, similar Sample Establishing unit 2 similar with foreign peoples, projection matrix set up unit 3, map unit 4 and diagnosis unit 5.

Wherein, data acquisition unit is for obtaining the gene expression data of test sample book and multiple training sample.And then similar Sample Establishing unit similar with foreign peoples, based on cosine similarity principle, determines the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample.

Projection matrix sets up unit according to described similar similar sample set sample set similar with described foreign peoples, sets up the projection matrix of low dimensional feature space according to presetting method.

It should be noted that, in other device embodiments of the present invention, projection matrix is set up unit and is specifically comprised:

Similar matrix construction unit between similar matrix and class in class, for according to similar similar sample sample similar with foreign peoples, to build in class similar matrix between similar matrix and class;

Wherein:

similar matrix in representation class, similar matrix between representation class, x ' _ithe gene expression data of training sample i, x ' _jrepresent the gene expression data of training sample j, represent the similar similar sample set of training sample j, the similar similar sample set of training sample i, represent the similar sample set of foreign peoples of training sample j, represent the similar sample set of foreign peoples of training sample i;

To build between class local Scatter Matrix construction unit in local Scatter Matrix and class, for according to similar matrix between similar matrix and class in class, to build between class local Scatter Matrix in local Scatter Matrix and class;

Wherein, S _w=X (D ^w-F ^w) X ^t, S _b=X (D ^b-F ^b) X ^t;

S _wlocal Scatter Matrix between representation class, S _blocal Scatter Matrix between representation class, X represents the gene expression data of sample, F ^bsimilar matrix between representation class, F ^wsimilar matrix in representation class, D ^wand D ^brepresent diagonal matrix,

Feature decomposition unit, for carrying out feature decomposition to local Scatter Matrix in local Scatter Matrix between described class and described class, ensureing that the ratio of between class distance and inter-object distance maximizes, obtaining several eigenwerts according to order from big to small;

Projection matrix sets up subelement, for setting up the projection matrix of low dimensional feature space according to eigenwert characteristic of correspondence vector.

Map unit utilizes described projection matrix that test sample book and multiple training sample are mapped to low dimensional feature space.Diagnosis unit determines the training sample nearest with test sample book in low dimensional feature space, gives test sample book, to complete the diagnosis to test sample book by the classification of this training sample.

Optionally, in order to reduce computation complexity, this device also comprises in other embodiments of the invention: the first data processing unit and the second data processing unit.

Wherein, data processing unit, extracts process and characteristic value normalization process for carrying out random parameters to the gene expression data of test sample book and multiple training sample.

Preferably, also comprise: the first data processing unit is used for carrying out random parameters to the gene expression data of test sample book and extracts process and characteristic value normalization process.

It should be noted that, a kind of methods for the diagnosis of diseases disclosed by the invention can be applicable to the diagnosis of breast cancer relapse.

Show that diagnosis of the present invention raises along with dimension by experiment always, tie up left and right 40, the present invention has surmounted the diagnosis that digraph differentiates neighbour's embedded mobile GIS, and can obtain excellent diagnostics result.

Finally, also it should be noted that, in this article, the such as relational terms of first and second grades and so on is only used for an entity or operation to separate with another entity or operational zone, and not necessarily requires or imply the relation that there is any this reality between these entities or operation or sequentially.And, term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability, thus make to comprise the process of a series of key element, method, article or equipment and not only comprise those key elements, but also comprise other key elements clearly do not listed, or also comprise by the intrinsic key element of this process, method, article or equipment.When not more restrictions, the key element limited by statement " comprising ... ", and be not precluded within process, method, article or the equipment comprising described key element and also there is other identical element.

In this instructions, each embodiment adopts the mode of going forward one by one to describe, and what each embodiment stressed is the difference with other embodiments, between each embodiment identical similar portion mutually see.

To the above-mentioned explanation of the disclosed embodiments, professional and technical personnel in the field are realized or uses the application.To be apparent for those skilled in the art to the multiple amendment of these embodiments, General Principle as defined herein when not departing from the spirit or scope of the application, can realize in other embodiments.Therefore, the application can not be restricted to these embodiments shown in this article, but will meet the widest scope consistent with principle disclosed herein and features of novelty.

Claims (6)

1. a methods for the diagnosis of diseases, is characterized in that, comprising:

Obtain the gene expression data of test sample book and multiple training sample;

Based on cosine similarity principle, determine the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample;

According to described similar similar sample set sample set similar with described foreign peoples, set up the projection matrix of low dimensional feature space according to presetting method;

Utilize described projection matrix that multiple training sample after test sample book after process and process is mapped to low dimensional feature space;

In low dimensional feature space, determine the training sample nearest with test sample book, give test sample book, to complete the diagnosis to test sample book by the classification of this training sample.

2. method according to claim 1, is characterized in that, described based on cosine similarity principle, determines the similar similar sample set of any one training sample sample set similar with foreign peoples, also comprise before according to the gene expression data of multiple training sample:

Random parameters extraction process and characteristic value normalization process are carried out to the gene expression data of test sample book and multiple training sample.

3. method according to claim 2, is characterized in that, described according to described similar similar sample set sample set similar with described foreign peoples, sets up the projection matrix of low dimensional feature space, comprising according to presetting method:

According to similar similar sample sample similar with foreign peoples, similar matrix between similar matrix and class in structure class;

Wherein:

similar matrix in representation class, similar matrix between representation class, x ' _ithe gene expression data of the training sample i after process, x ' _jrepresent the gene expression data of the rear training sample j of process, represent the similar similar sample set of training sample j, the similar similar sample set of training sample i, represent the similar sample set of foreign peoples of training sample j, represent the similar sample set of foreign peoples of training sample i;

According to similar matrix between similar matrix and class in class, to build between class local Scatter Matrix in local Scatter Matrix and class;

Wherein, S _w=X (D ^w-F ^w) X ^t, S _b=X (D ^b-F ^b) X ^t;

S _wlocal Scatter Matrix in representation class, S _blocal Scatter Matrix between representation class, X represents the gene expression data of sample, F ^wsimilar matrix in representation class, F ^bsimilar matrix between representation class, D ^wand D ^brepresent diagonal matrix, $D_{ii}^w={\mathrm{\Σ}}_j{F}_{ij}^w,D_{ii}^b={\mathrm{\Σ}}_j{F}_{ij}^b;$

Feature decomposition is carried out to local Scatter Matrix in local Scatter Matrix between described class and described class, ensures that the ratio of between class distance and inter-object distance maximizes, obtain several eigenwerts according to order from big to small;

The projection matrix of low dimensional feature space is set up according to eigenwert characteristic of correspondence vector.

4. a medical diagnosis on disease device, is characterized in that, comprising:

Data acquisition unit, for obtaining the gene expression data of test sample book and multiple training sample;

Similar Sample Establishing unit similar with foreign peoples, for based on cosine similarity principle, determines the similar similar sample set of any one training sample sample set similar with foreign peoples according to the gene expression data of multiple training sample;

Projection matrix sets up unit, for according to described similar similar sample set sample set similar with described foreign peoples, sets up the projection matrix of low dimensional feature space according to presetting method;

Map unit, is mapped to low dimensional feature space for utilizing described projection matrix by test sample book and multiple training sample;

Diagnosis unit, in low dimensional feature space, determines the training sample nearest with test sample book, gives test sample book, to complete the diagnosis to test sample book by the classification of this training sample.

5. device according to claim 4, is characterized in that, also comprises: data processing unit, extracts process and characteristic value normalization process for carrying out random parameters to the gene expression data of test sample book and multiple training sample.

6. device according to claim 4, is characterized in that, described projection matrix is set up unit and comprised:

Similar matrix construction unit between similar matrix and class in class, for according to similar similar sample sample similar with foreign peoples, to build in class similar matrix between similar matrix and class;

Wherein:

similar matrix in representation class, similar matrix between representation class, x ' _ithe gene expression data of the training sample i after process, x ' _jthe gene expression data of training sample j is represented after process, represent the similar similar sample set of training sample j, the similar similar sample set of training sample i, represent the similar sample set of foreign peoples of training sample j, represent the similar sample set of foreign peoples of training sample i;

To build between class local Scatter Matrix construction unit in local Scatter Matrix and class, for according to similar matrix between similar matrix and class in class, to build between class local Scatter Matrix in local Scatter Matrix and class;

Wherein, S _w=X (D ^w-F ^w) X ^t, S _b=X (D ^b-F ^b) X ^t;

S _wlocal Scatter Matrix in representation class, S _blocal Scatter Matrix between representation class, X represents the gene expression data of sample, F ^wsimilar matrix in representation class, F ^bsimilar matrix between representation class, D ^wand D ^brepresent diagonal matrix, $D_{ii}^w={\mathrm{\Σ}}_j-F_{ij}^w,D_{ii}^b={\mathrm{\Σ}}_j{F}_{ij}^b;$

Feature decomposition unit, for carrying out feature decomposition to local Scatter Matrix in local Scatter Matrix between described class and described class, ensureing that the ratio of between class distance and inter-object distance maximizes, obtaining several eigenwerts according to order from big to small;

Projection matrix sets up subelement, for setting up the projection matrix of low dimensional feature space according to eigenwert characteristic of correspondence vector.