CN107731283A - A kind of image radio system based on more subspace modelings - Google Patents
A kind of image radio system based on more subspace modelings Download PDFInfo
- Publication number
- CN107731283A CN107731283A CN201710992065.2A CN201710992065A CN107731283A CN 107731283 A CN107731283 A CN 107731283A CN 201710992065 A CN201710992065 A CN 201710992065A CN 107731283 A CN107731283 A CN 107731283A
- Authority
- CN
- China
- Prior art keywords
- image
- hypergraph
- magnetic resonance
- mrow
- mtd
- 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.)
- Pending
Links
- 238000005481 NMR spectroscopy Methods 0.000 claims abstract description 22
- 210000004556 brain Anatomy 0.000 claims abstract description 20
- 238000010276 construction Methods 0.000 claims abstract description 19
- 208000010877 cognitive disease Diseases 0.000 claims abstract description 16
- 208000027061 mild cognitive impairment Diseases 0.000 claims abstract description 14
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 238000003384 imaging method Methods 0.000 claims description 30
- 239000011159 matrix material Substances 0.000 claims description 23
- 239000013598 vector Substances 0.000 claims description 15
- 238000002598 diffusion tensor imaging Methods 0.000 claims description 14
- 238000000034 method Methods 0.000 claims description 14
- 238000002599 functional magnetic resonance imaging Methods 0.000 claims description 10
- 238000002372 labelling Methods 0.000 claims description 9
- 230000010412 perfusion Effects 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 6
- 230000000284 resting effect Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 208000024827 Alzheimer disease Diseases 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000002595 magnetic resonance imaging Methods 0.000 description 5
- 230000000052 comparative effect Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000012216 screening Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 2
- 208000015122 neurodegenerative disease Diseases 0.000 description 2
- 239000004576 sand Substances 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 206010012289 Dementia Diseases 0.000 description 1
- 206010039966 Senile dementia Diseases 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000003376 axonal effect Effects 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 210000005013 brain tissue Anatomy 0.000 description 1
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 1
- 230000006999 cognitive decline Effects 0.000 description 1
- 239000002872 contrast media Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005111 flow chemistry technique Methods 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 238000002610 neuroimaging Methods 0.000 description 1
- 231100000915 pathological change Toxicity 0.000 description 1
- 230000036285 pathological change Effects 0.000 description 1
- 230000007170 pathology Effects 0.000 description 1
- 238000002600 positron emission tomography Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 210000001519 tissue Anatomy 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a kind of image radio system based on more subspace modelings, described image radio system includes：Nuclear magnetic resonance image collecting device, image preprocess apparatus, feature deriving means, hypergraph construction device, image preselector, nuclear magnetic resonance image collecting device is used for the brain nuclear magnetic resonance image for gathering target group, and the image includes many reference amounts data of the brain nuclear magnetic resonance image of mild cognitive impairment individual and the brain nuclear magnetic resonance image of normal individual；Described image radio system is handled image, region of interesting extraction and hypergraph build, and carry out image pre-selection based on constructed hypergraph.
Description
Technical Field
The invention belongs to the field of computer-aided image processing equipment, and particularly relates to a system for pre-selecting images by using a multi-subspace joint modeling method.
Background
Alzheimer's Disease (AD), commonly known as senile dementia, is a neurodegenerative disorder. Related medical studies have shown that an intermediate state between normal aging and dementia exists, called Mild Cognitive Impairment (MCI), a Cognitive disorder syndrome. Compared to normal elderly people of age and education, patients had mild cognitive decline, but daily abilities were not significantly affected. The diagnosis is effectively carried out in the stage of mild cognitive impairment, and the targeted treatment is carried out to delay the pathological changes, thereby having very important significance for improving the life quality of patients and other families. Therefore, mild cognitive impairment is diagnosed before the onset of alzheimer's disease, attracting the attention of many researchers.
Recent studies have shown that the brain undergoes structural and functional changes before the pathology switches to alzheimer's disease, a property that can be used to identify mild cognitive impairment. It is now common in medicine to examine brain-specific imaging performance using neuroimaging methods. Here, Magnetic Resonance Imaging (MRI) which is more sensitive to the change of the blood vessels under the cortex and the change of the special structure is used to scan the brain structure, and MRI has the characteristics of a plurality of Imaging parameters, namely T1 weighted Imaging (T1), T2 weighted Imaging (T2), Diffusion Tensor Imaging (DTI), resting-state functional MRI (rs-fMRI) and the like. They are used clinically for routine scanning of structural and functional features of the brain. For example, T1 weighted imaging can detect brain tissue type information, magnetic resonance diffusion tensor imaging can detect macroscopic axonal tissue in the nervous system, and resting state functional magnetic resonance provides regional interactions when an individual lacks clear tasks. As a related novel technique, arterial spin labeling perfusion (ASL) is used as a measurement of brain perfusion, and does not require injection of contrast agent and other measures. Recent studies have shown that arterial spin labeling can achieve effective prediction of early neurodegenerative disorders. In clinical treatment, multi-parameter mri is data that is relatively easy to obtain simultaneously.
Recent research results show that researchers have done much work in combination with multi-modality (parametric) data to improve the diagnostic accuracy, such as in combination with magnetic resonance, positron emission tomography, cerebrospinal fluid images, and the like. However, in most of the works, the correlation model between the samples is established independently for each modality, so that important correlation information existing between the modalities is ignored. Indeed, fusing multi-modal information is a challenging task, as the relationship between samples may be different in different modalities.
Therefore, a system capable of effectively fusing multiple modalities is needed to pre-select brain images of a patient so as to assist a doctor in performing previous-stage image and patient screening and reduce the workload of the doctor.
Disclosure of Invention
The invention aims to overcome the defects of the existing method and provides a system which can utilize multi-modal nuclear magnetic resonance image data to provide better high-order association modeling processing for association information among the artifacts so as to realize the preselection of the mild cognitive impairment images.
Specifically, the invention provides an image preselection system based on multi-subspace modeling, which is characterized in that the image preselection system comprises: a nuclear magnetic resonance image acquisition device, an image preprocessing device, a feature extraction device, a hypergraph construction device and an image preselection device,
the nuclear magnetic resonance image acquisition equipment is used for acquiring brain nuclear magnetic resonance images of target people, and the images comprise brain nuclear magnetic resonance images of mild cognitive impairment individuals and multi-parameter data of brain nuclear magnetic resonance images of normal individuals;
the image preprocessing device divides the brain region of interest of the multi-parameter nuclear magnetic resonance image;
the feature extraction device extracts required features according to the characteristics of each divided interest region, and performs quantitative calculation on the features in the interest regions to obtain required feature vectors; the hypergraph construction device establishes a relation between samples by constructing a hypergraph structure, wherein in the hypergraph, an Euclidean space is represented by K-neighbor connection, and a reconstruction space of a feature is represented by sparse representation of the feature;
and the image preselection device maps the target image into the hypergraph for image preselection by utilizing the same rule based on the constructed hypergraph.
Preferably, the imaging data includes imaging data of four parameters of T1 weighted imaging, magnetic resonance diffusion tensor imaging, resting state functional magnetic resonance imaging and magnetic resonance arterial spin labeling perfusion imaging.
Preferably, the feature extraction device generates a 180-dimensional feature vector for the T1 image data of each individual, and generates a 90-dimensional feature vector for the magnetic resonance diffusion tensor imaging, the resting state functional magnetic resonance imaging and the magnetic resonance arterial spin labeling perfusion imaging of each individual.
Preferably, in the hypergraph constructed by the hypergraph construction apparatus, each hypergraph includes a plurality of vertices, each vertex is used as a center to establish a hypergraph, g ═ represents a hypergraph structure, v represents a set of vertices, and each vertex represents a sample point; epsilon represents the set of super-edges and W represents the diagonal matrix formed by the weights of the super-edges, a correlation matrix H is generated from the constructed hypergraph structure,
wherein eiIs represented by viThe generated super-edge is generated by the method,is shown in reconstruction viThe jth element in the linear combination coefficient x.
Preferably, the sparse representation comprises: let v and P represent the feature vectors of the central sample point and k nearest neighbor points, respectively, and solve the following function to find the coefficients of the sparse representation of P to v:
argminx{||Px-v||2+γ||x||1}，
in the formula, the first term represents reconstruction of k neighboring points to the center point, the second term is a sparse penalty term, and γ is a trade-off parameter.
Preferably, the hypergraph construction means represents the manifold space by geodetic neighbour connections.
The system of the present invention provides full flow processing from image acquisition to image processing and final image screening.
The system can estimate the associated information among samples in a plurality of subspaces, and the basic idea is that the associated information of the similarity of different samples is estimated by combining the image data information of a plurality of modes under a semi-supervision method, namely the possibility that different samples belong to the same category is estimated, then the associated information is expressed on a plurality of characteristic subspaces in a hypergraph structure (comprising training data and test data), and then the hypergraph structures of the plurality of modes and the subspaces are combined, and the classification of the test data is realized by using a hypergraph learning method, namely the prediction of mild cognitive impairment is realized.
Compared with other Mild Cognitive Impairment (MCI) image preselection systems, the system provided by the invention starts from a plurality of subspaces and mines higher-order associated information among sample points, so that the obtained information is more comprehensive. Compared with other hypergraph construction systems, the hypergraph construction system has the advantages that a hypergraph structure with better performance can be built, and contained information is richer and more effective.
Drawings
Fig. 1 is a schematic diagram of the working process of an image preselection system based on multi-subspace modeling.
Fig. 2 shows the euclidean subspace.
Fig. 3 shows the construction of a super edge.
Fig. 4 is characteristic dimension information of different image data.
FIG. 5 is a comparative plot of the results of T1 parametric data.
FIG. 6 is a comparative view of the results of DTI parametric data.
FIG. 7 is a comparative plot of the results of RS-fMRI data.
Fig. 8 is a comparative view of ASL data results.
Figure 9 multi-parametric imaging data results.
Detailed Description
The invention is described in detail below with reference to the drawings and the embodiments thereof, but the scope of the invention is not limited thereto.
In order to achieve the purpose of the invention, the invention provides a multi-subspace-based MCI image preselection system.
The image preselection system comprises: the system comprises nuclear magnetic resonance image acquisition equipment, an image preprocessing device, a feature extraction device, a hypergraph construction device and an image preselection device.
Image acquisition
The nuclear magnetic resonance image acquisition equipment is used for acquiring brain nuclear magnetic resonance images of target people. MRI examinations are characterized by a number of imaging parameters, namely T1-weighted imaging (T1), T2-weighted imaging (T2), magnetic resonance Diffusion Tensor Imaging (DTI), and so on. After image data is obtained, screening the data, and after primary screening, selecting 80 groups of mild cognitive impairment individuals and 80 groups of normal and non-diseased control data. Here we use the Imaging data of four parameters, T1 weighted Imaging (T1-weighted, T1), magnetic resonance Diffusion Tensor Imaging (DTI), Resting-State functional magnetic resonance Imaging (RS-fMRI) and magnetic resonance Arterial Spin labeling perfusion Imaging (ASL). Each set of data contains magnetic imaging data for the four parameters.
And the image preprocessing device is used for dividing the brain interested region of the multi-parameter nuclear magnetic resonance image. The feature extraction device extracts required features according to the characteristics of each divided interest region, and performs quantitative calculation on the features in the interest region to obtain required feature vectors. According to the anatomical knowledge, the area on the brain can be divided into 90 interest areas, and the required features can be extracted according to the characteristics of each divided interest area. The image data of the four parameters are divided by a deformation registration algorithm according to the positions of the 90 interest areas, the image data of each parameter is divided into 90 areas, and the required feature vectors can be obtained by quantitatively calculating some features reflecting medical characteristics in the 90 interest areas. Thus, each individual T1 image data may generate a 180-dimensional feature vector, and the DTI, RS-fMRI, and ASL image data may generate a 90-dimensional feature vector, respectively, as shown in fig. 2.
The hypergraph construction apparatus establishes a relationship between samples by constructing a hypergraph structure over multiple subspaces. While the traditional hypergraph model is constructed by relying on K Nearest Neighbor (KNN), the model proposed by us relies on multi-subspace information, including expressing euclidean space by using K nearest neighbor connections, expressing reconstructed space of features by using sparse representation of features, and expressing manifold space by using geodesic nearest neighbor connections.
The European subspace: the conventional hypergraph structure is built in the euclidean space as shown in fig. 2. Each point is a vertex, and each vertex represents a feature vector of one sample point. According to the conventional graph theory, each edge can only include two vertices, and in the hypergraph structure, each hyperedge can include a number of vertices, and the number of included vertices are determined by a parameter K of the K-nearest neighbor algorithm. I establish a super edge by taking each vertex as a center, and assuming that N vertices exist, N super edges are generated. Let g ═ v, epsilon, W to represent a hypergraph structure, v representing a set of vertices, each vertex representing a sample point; ε represents the set of super-edges, and W represents the diagonal matrix made up of the weights of the super-edges. Then, we can generate a correlation matrix H from the constructed hypergraph structure, that is:
feature sparse subspace: the K-nearest neighbor algorithm constructs a hypergraph in an Euclidean space, which is susceptible to noise and extreme values in performance, and if the data has noise, the robustness of the model is poor. Thus, in the system of the present invention, sparse representations of features are used to establish relationships between samples. The method comprises the steps of enabling a linear combination of k adjacent points of a sample point in Euclidean space to represent the sample point as much as possible, and simultaneously requiring coefficients of the linear combination to be as sparse as possible, so as to mine the points which are most relevant to a central point from the k adjacent points in Euclidean space and eliminate the points which are least relevant to the central point, and further extracting the relevant information existing between the sample points. In representation, let v and P represent the feature vectors of the central sample point and the k nearest neighbors, respectively. We find the coefficients of the sparse representation of P over v by solving the following function:
argminx{||Px-v||2+γ||x||1}，
the first term of the above equation represents the reconstruction of the center point from k neighboring points in order to minimize the error of linear combination, and the second term is the sparse penaltyThe objective is to make the coefficients x as sparse as possible. Gamma is a weighting parameter. Constraint xiThe reason for being more than or equal to 0 is to ensure that the coefficient of the linear combination is not negative, and the neighbor points do not have adverse effects on the central point. By this method, the incidence matrix of the hypergraph can be obtained as follows:
wherein v isiIs represented by viThe generated super-edge is generated by the method,is shown in reconstruction viThe jth element in the coefficient x.
A manifold space: the hypergraph constructed in the original feature space may not be optimal, and the system of the present invention estimates the relationships between samples in the manifold space. First, the hypergraph construction apparatus maps all sample points into a manifold structure, and then constructs a hypergraph in this manifold space. Considering the superiority of the hypergraph structure in researching high-order association of data, the hypergraph construction device adopts a double-layer graph learning algorithm based on the hypergraph which is newly designed by the invention. Firstly, mapping features into manifold subspace by using a semi-supervised learning algorithm of a graph to estimate the relation between samples, and then sending the obtained associated information into a hypergraph model to predict a result.
More specifically, the hypergraph construction apparatus constructs a graph structure using all the sample points, and using the vertex vsRepresents a sample point, ε (v)s，vt) Represents an edge, represents vsAnd vtThe similarity of (a) is defined as follows:
wherein d is2(vs，vt) Represents vsAnd vtSquared euclidean distance of (d). At the same time, let D (s, s) be ∑tW (s, t). According to the existing graph learning algorithm, the relationship matrix between the vertexes can be obtained as follows:
wherein S ═ D-1/2WD-1/2ζ is a trade-off parameter.
The hypergraph structure is then constructed using the resulting relationship matrix R. And selecting each vertex as a central point, and selecting k most relevant vertexes thereof according to the relation between the vertexes in the R matrix to connect with the vertexes to form a hyper-edge. And when v ∈ e incidence matrix H (v, e) is 1, otherwise, the incidence matrix is 0. Each super edge has a weight of
The feature vector of each parameter is subjected to feature processing and hypergraph construction in the three steps. The following hypergraph construction device fuses the hypergraphs of the plurality of subspaces to generate a total hypergraph, and the fusing method comprises the following steps: for each parameter data, when constructing a super edge with a vertex as a center point, the above three super edges corresponding to the center point constructed in the subspace and the weighted union set are merged into a total super edge with the vertices occurring in the three super edges, as shown in fig. 3. The correlation matrix H can be generated from the resulting total hypergraph. Placing the H matrices of these four parameters side by side together results in H which is finally fed to the classification step, i.e. H ═ HT1HDTIHRS-fMRIHASL]. The system adopts Imaging data of multiple parameters, namely T1 weighted Imaging (T1-weighted, T1), magnetic resonance Diffusion Tensor Imaging (DTI), Resting-State functional magnetic resonance Imaging (RS-fMRI) and magnetic resonance dynamic MRI (dynamic magnetic resonance Imaging)Pulse spin labeling perfusion imaging (ASL), each imaging data can generate a corresponding correlation matrix H by the above method. T1 weighted imaging (T1) data generating correlation matrix HT1Magnetic resonance diffusion tensor imaging (D)TI) Data generation incidence matrix HDTIThe resting-state functional magnetic resonance imaging (RS-fMRI) data generates a correlation matrix HRS-fMRIMagnetic resonance arterial spin labeling perfusion imaging (ASL) data generation correlation matrix HASLI.e., the different matrices therein represent the respective correlation matrices H generated from the different parametric imaging data, respectively.
Having obtained the correlation matrix H, we compute the order of each vertex V ∈ V and each hyperedge e ∈ ε, i.e., d (V) ∑ Σe∈εw (e) h (v, e) and δ (e) ═ Σv∈Vh (v, e). According to a hypergraph learning algorithm, solving the following objective function to obtain the association between the sample points and the labels:
argminF{Ω(F)+λRemp(F)} (6)
the first term is a regularization term of the hypergraph, the second term is an empirical loss term, and λ is a trade-off parameter. Ω is defined as:
wherein,the empirical loss term is defined as:
whereinIs a label matrix, when the sample point viIn the k-th class, Y (i, k) is 1, and others are 0.
By making the partial derivative of the objective function to F0, the solution to the objective function can be obtained as:
where F is the confidence score, vertex viBelongs to the class with the largest score value in F (i, i). The fractional matrix F is a matrix with dimension n × c, wherein n is the number of samples, and c is the number of classes. Consider that the ith sample corresponds to the ith row of the F matrix, i.e., F (I,: the number of columns j corresponding to the largest item in F (I,: I) is the classification result, i.e., the ith sample image is classified into the jth class.
The foregoing is considered as illustrative and not restrictive, and the appended claims are intended to cover all such modifications, changes, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims.
While the principles of the invention have been described in detail in connection with the preferred embodiments thereof, it will be understood by those skilled in the art that the foregoing embodiments are merely illustrative of exemplary implementations of the invention and are not limiting of the scope of the invention. The details of the embodiments are not to be interpreted as limiting the scope of the invention, and any obvious changes, such as equivalent alterations, simple substitutions and the like, based on the technical solution of the invention, can be interpreted without departing from the spirit and scope of the invention.
Claims (6)
1. An image preselection system based on multi-subspace modeling, the image preselection system comprising: a nuclear magnetic resonance image acquisition device, an image preprocessing device, a feature extraction device, a hypergraph construction device and an image preselection device,
the nuclear magnetic resonance image acquisition equipment is used for acquiring brain nuclear magnetic resonance images of target people, and the images comprise brain nuclear magnetic resonance images of mild cognitive impairment individuals and multi-parameter data of brain nuclear magnetic resonance images of normal individuals;
the image preprocessing device divides the brain region of interest of the multi-parameter nuclear magnetic resonance image;
the feature extraction device extracts required features according to the characteristics of each divided interest region, and performs quantitative calculation on the features in the interest regions to obtain required feature vectors;
the hypergraph construction device establishes a relation between samples by constructing a hypergraph structure, wherein in the hypergraph, an Euclidean space is represented by K-neighbor connection, and a reconstruction space of a feature is represented by sparse representation of the feature;
and the image preselection device maps the target image into the hypergraph for image preselection by utilizing the same rule based on the constructed hypergraph.
2. The multi-subspace modeling based image preselection system of claim 1, wherein said imaging data includes imaging data for four parameters of T1 weighted imaging, magnetic resonance diffusion tensor imaging, resting state functional magnetic resonance imaging, and magnetic resonance arterial spin labeling perfusion imaging.
3. The system of claim 1, wherein the feature extraction means generates a 180-dimensional feature vector for the T1 image data of each individual, and generates a 90-dimensional feature vector for the magnetic resonance diffusion tensor imaging, the resting state functional magnetic resonance imaging, and the magnetic resonance arterial spin labeling perfusion imaging of each individual.
4. The image preselection system based on multi-subspace modeling as claimed in claim 1, wherein said hypergraph construction means constructs a hypergraph in which each hypergraph comprises a plurality of vertices, each vertex is used as a center to construct a hypergraph, and the order of each hypergraph is thatTo represent a hypergraph structure, v represents a set of vertices, each vertex representing a vertexTable one sample point; ε represents the set of super-edges, and W represents the diagonal matrix made up of the weights of the super-edges, a correlation matrix H is generated from the constructed hypergraph structure,
<mrow> <mi>H</mi> <mrow> <mo>(</mo> <mi>v</mi> <mo>,</mo> <mi>e</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mn>1</mn> </mtd> <mtd> <mrow> <mi>i</mi> <mi>f</mi> </mrow> </mtd> <mtd> <mrow> <mi>v</mi> <mo>&Element;</mo> <mi>e</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mrow> <mi>i</mi> <mi>f</mi> </mrow> </mtd> <mtd> <mrow> <mi>v</mi> <mo>&NotElement;</mo> <mi>e</mi> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow>
wherein eiIs represented by viThe generated super-edge is generated by the method,is shown in reconstruction viThe jth element in the coefficient x.
5. The multi-subspace modeling based image preselection system of claim 1, wherein the sparse representation comprises: let v and P represent the feature vectors of the central sample point and k nearest neighbors, respectively, and solve the following function to find the coefficients of the sparse representation of P to v:
argminx{||Px-v||2+γ||x||1}，
<mrow> <mi>s</mi> <mo>.</mo> <mi>t</mi> <mo>.</mo> <mo>&ForAll;</mo> <mi>i</mi> <mo>,</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>&GreaterEqual;</mo> <mn>0</mn> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>
in the formula, the first term represents reconstruction of k neighboring points to the center point, the second term is a sparse penalty term, and γ is a trade-off parameter.
6. The system of claim 1, wherein said hypergraph construction means represents manifold space by geodetic neighbor connections.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710992065.2A CN107731283A (en) | 2017-10-23 | 2017-10-23 | A kind of image radio system based on more subspace modelings |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710992065.2A CN107731283A (en) | 2017-10-23 | 2017-10-23 | A kind of image radio system based on more subspace modelings |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107731283A true CN107731283A (en) | 2018-02-23 |
Family
ID=61213133
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710992065.2A Pending CN107731283A (en) | 2017-10-23 | 2017-10-23 | A kind of image radio system based on more subspace modelings |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107731283A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020000721A1 (en) * | 2018-06-29 | 2020-01-02 | 平安科技（深圳）有限公司 | Disease test method, test server, and computer readable storage medium |
| CN110895700A (en) * | 2018-09-12 | 2020-03-20 | 北京京东尚科信息技术有限公司 | Image recognition method and system |
| CN112735570A (en) * | 2021-01-09 | 2021-04-30 | 深圳先进技术研究院 | Image-driven brain atlas construction method, device, equipment and storage medium |
| CN113724857A (en) * | 2021-08-27 | 2021-11-30 | 清华大学深圳国际研究生院 | Automatic diagnosis device for eye ground disease based on eye ground image retina blood vessel |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024036A (en) * | 2010-11-29 | 2011-04-20 | 清华大学 | Three-dimensional object retrieval method and device based on hypergraphs |
| US20140343399A1 (en) * | 2013-05-20 | 2014-11-20 | Stefan Posse | Seed-Based Connectivity Analysis in Functional MRI |
| EP2889743A1 (en) * | 2013-12-23 | 2015-07-01 | Samsung Electronics Co., Ltd | Method of and apparatus for providing medical image |
| CN105808732A (en) * | 2016-03-10 | 2016-07-27 | 北京大学 | Integration target attribute identification and precise retrieval method based on depth measurement learning |
| CN106875401A (en) * | 2017-01-10 | 2017-06-20 | 中国科学院深圳先进技术研究院 | The analysis method of multi-modal image group, device and terminal |
-
2017
- 2017-10-23 CN CN201710992065.2A patent/CN107731283A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024036A (en) * | 2010-11-29 | 2011-04-20 | 清华大学 | Three-dimensional object retrieval method and device based on hypergraphs |
| US20140343399A1 (en) * | 2013-05-20 | 2014-11-20 | Stefan Posse | Seed-Based Connectivity Analysis in Functional MRI |
| EP2889743A1 (en) * | 2013-12-23 | 2015-07-01 | Samsung Electronics Co., Ltd | Method of and apparatus for providing medical image |
| CN105808732A (en) * | 2016-03-10 | 2016-07-27 | 北京大学 | Integration target attribute identification and precise retrieval method based on depth measurement learning |
| CN106875401A (en) * | 2017-01-10 | 2017-06-20 | 中国科学院深圳先进技术研究院 | The analysis method of multi-modal image group, device and terminal |
Non-Patent Citations (1)
| Title |
|---|
| 彭瑶等: "基于超图的多模态特征选择算法及其应用", 《计算机科学与探索》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020000721A1 (en) * | 2018-06-29 | 2020-01-02 | 平安科技（深圳）有限公司 | Disease test method, test server, and computer readable storage medium |
| CN110895700A (en) * | 2018-09-12 | 2020-03-20 | 北京京东尚科信息技术有限公司 | Image recognition method and system |
| CN112735570A (en) * | 2021-01-09 | 2021-04-30 | 深圳先进技术研究院 | Image-driven brain atlas construction method, device, equipment and storage medium |
| CN112735570B (en) * | 2021-01-09 | 2022-02-22 | 深圳先进技术研究院 | Image-driven brain atlas construction method, device, equipment and storage medium |
| WO2022147871A1 (en) * | 2021-01-09 | 2022-07-14 | 深圳先进技术研究院 | Image-driven brain map construction method and apparatus, device, and storage medium |
| CN113724857A (en) * | 2021-08-27 | 2021-11-30 | 清华大学深圳国际研究生院 | Automatic diagnosis device for eye ground disease based on eye ground image retina blood vessel |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lei et al. | Self-calibrated brain network estimation and joint non-convex multi-task learning for identification of early Alzheimer's disease | |
| Liu et al. | Weakly supervised deep learning for brain disease prognosis using MRI and incomplete clinical scores | |
| CN113040715B (en) | Human brain function network classification method based on convolutional neural network | |
| CN113571195B (en) | Early Alzheimer disease prediction model based on cerebellar function connection characteristics | |
| CN107944490B (en) | Image classification method based on semi-multimodal fusion feature reduction framework | |
| Yang et al. | A deep neural network study of the ABIDE repository on autism spectrum classification | |
| Song et al. | A modified probabilistic neural network for partial volume segmentation in brain MR image | |
| CN113616184A (en) | Brain network modeling and individual prediction method based on multi-mode magnetic resonance image | |
| WO2021103287A1 (en) | Brain disease progression prediction method and system based on weak supervision multi-task matrix completion | |
| Gore et al. | Comparative study of various techniques using deep Learning for brain tumor detection | |
| CN111488914A (en) | Alzheimer disease classification and prediction system based on multitask learning | |
| CN106023194A (en) | Amygdaloid nucleus spectral clustering segmentation method based on resting state function connection | |
| CN107731283A (en) | A kind of image radio system based on more subspace modelings | |
| Klein et al. | Early diagnosis of dementia based on intersubject whole-brain dissimilarities | |
| Jie et al. | Brain connectivity hyper-network for MCI classification | |
| Spetsieris et al. | Spectral guided sparse inverse covariance estimation of metabolic networks in Parkinson's disease | |
| Ye et al. | Regional manifold learning for disease classification | |
| Glaister et al. | Thalamus segmentation using multi-modal feature classification: Validation and pilot study of an age-matched cohort | |
| CN108596228B (en) | Brain function magnetic resonance image classification method based on unsupervised fuzzy system | |
| CN112233086A (en) | fMRI data classification and identification method and device based on brain region functional connection | |
| CN112233805A (en) | Mining method for biomarkers based on multi-map neuroimaging data | |
| Sørensen et al. | Dementia diagnosis using MRI cortical thickness, shape, texture, and volumetry | |
| Bloy et al. | White matter atlas generation using HARDI based automated parcellation | |
| Xi et al. | Brain Functional Networks with Dynamic Hypergraph Manifold Regularization for Classification of End-Stage Renal Disease Associated with Mild Cognitive Impairment. | |
| Pallawi et al. | Study of Alzheimer’s disease brain impairment and methods for its early diagnosis: a comprehensive survey |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180223 |
|
| RJ01 | Rejection of invention patent application after publication |