CN107680677A - Neuropsychiatric disease sorting technique based on brain network analysis - Google Patents
Neuropsychiatric disease sorting technique based on brain network analysis Download PDFInfo
- Publication number
- CN107680677A CN107680677A CN201710940330.2A CN201710940330A CN107680677A CN 107680677 A CN107680677 A CN 107680677A CN 201710940330 A CN201710940330 A CN 201710940330A CN 107680677 A CN107680677 A CN 107680677A
- Authority
- CN
- China
- Prior art keywords
- sample
- mrow
- network
- brain
- brain network
- 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
- 210000004556 brain Anatomy 0.000 title claims abstract description 261
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 50
- 201000010099 disease Diseases 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000003012 network analysis Methods 0.000 title claims abstract description 18
- 239000000835 fiber Substances 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 12
- 210000003710 cerebral cortex Anatomy 0.000 claims description 5
- 230000002739 subcortical effect Effects 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims 1
- 239000002245 particle Substances 0.000 claims 1
- 230000007170 pathology Effects 0.000 abstract description 6
- 230000008569 process Effects 0.000 abstract description 3
- 230000004927 fusion Effects 0.000 abstract description 2
- 238000012512 characterization method Methods 0.000 abstract 1
- 230000002159 abnormal effect Effects 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 238000002599 functional magnetic resonance imaging Methods 0.000 description 6
- 210000004126 nerve fiber Anatomy 0.000 description 6
- 208000035475 disorder Diseases 0.000 description 5
- 208000024827 Alzheimer disease Diseases 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000001537 neural effect Effects 0.000 description 3
- 238000002610 neuroimaging Methods 0.000 description 3
- 208000020016 psychiatric disease Diseases 0.000 description 3
- 208000014644 Brain disease Diseases 0.000 description 2
- 238000003759 clinical diagnosis Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000007830 nerve conduction Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 206010010144 Completed suicide Diseases 0.000 description 1
- 241000845077 Iare Species 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 210000003484 anatomy Anatomy 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000003925 brain function Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 208000010877 cognitive disease Diseases 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 208000013403 hyperactivity Diseases 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 208000027061 mild cognitive impairment Diseases 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a kind of neuropsychiatric disease sorting technique based on brain network analysis, it is extremely complicated for brain network, sample number is relatively fewer, neuropsychiatric disease species various the problem of being difficult to Accurate classification, the application builds multilayer brain network first with region of interest and connection features parameter, then multilayer brain network is described using Local Clustering coefficient and the two local network structure features of node part importance scores and node significance score this global network architectural feature, the fusion of multilayer brain network characterization is finally carried out compared to the difference of normal sample using to be sorted and known classification samples, the distance between sample is calculated to distinguish different types of neuropsychiatric disease；The present processes can not only correctly be classified to known pathology, the clear and definite neuropsychiatric disease of feature, also to the not yet clear and definite neuropsychiatric disease of unknown pathology, feature or can be had the unknown neuropsychiatric disease progress effectively classification of similar features with known neuropsychiatric disease and be identified.
Description
Technical Field
The invention belongs to the field of photoelectric transmission, and particularly relates to an all-fiber distributed acoustic wave sensing technology.
Background
With modern brain imaging techniques, components in the brain and their connections can be described from multiple levels, such as nanoscale, microscale, and millimeter-scale, with neurons as objects, cortical-like functional pillars as objects, and brain regions as objects. (Zhang Xue, Liu Li, Guo Aike, "brain function connection atlas and brain-like intelligence research" leader research development and prospect, "Proc. Natl. Acad. Sci. China, 2016(7): 737-.
In recent years, the incidence of neuropsychiatric diseases such as alzheimer's disease and depression has been increasing. Statistically, over 530 million patients with neuropsychiatric diseases were enrolled in china at the beginning of 2017, which increased by 24% compared to 429 ten thousand enrolled in 2015 at the beginning. Meanwhile, neuropsychiatric diseases have attracted a wide range of social attention, such as suicide of young people caused by depression, a tendency of younger people to develop alzheimer's disease, and the like. How to accurately and effectively identify various neuropsychiatric diseases can help doctors to make a definite diagnosis in early stage of the diseases and delay the onset time of Alzheimer's disease and the like through medicines, or adopt corresponding means to perform targeted treatment in time according to the illness state of patients, thereby having important significance.
At present, the difficulty of neuropsychiatric disease classification based on brain network analysis lies mainly in:
1. the human brain is an extremely complex system, and it is very challenging how to accurately and comprehensively describe the brain network by using the brain network and extract relevant features to reflect changes which various neuropsychiatric diseases may bring to the brain network;
2. because neuropsychiatric diseases relate to privacy of patients, expensive manufacturing cost of brain imaging equipment and the like, the number of samples of complete neuropsychiatric disease patients which can be researched is relatively small, and the mainstream high-performance classification method usually needs a huge sample library, so that the neuropsychiatric diseases with small sample scale are difficult to be effectively classified by using the mainstream classification method;
3. neuropsychiatric disorders are of a wide variety, many disorders have similar characteristics, and it is challenging to accurately distinguish between different types of neuropsychiatric disorders by analyzing these characteristics.
At present, the method for classifying neuropsychiatric diseases by using brain network analysis at home and abroad can be mainly divided into two types:
1. method based on local features: this type of method generally combines the professional domain knowledge in medicine to perform comprehensive analysis of relevant areas in the brain network to classify neuropsychiatric diseases. For example, the fiber density of the brain in a certain region of the brain is checked by comparing a normal sample with an abnormal sample, the connection relationship between certain regions of interest of the brain is analyzed by comparing and the like. These classification methods can effectively identify neuropsychiatric diseases with known pathology and definite characteristics, but cannot accurately classify neuropsychiatric diseases with unknown pathology and definite characteristics. Also, such methods may misclassify certain unknown neuropsychiatric diseases as known neuropsychiatric diseases with similar characteristics.
2. The method based on the single connection relation comprises the following steps: the classification method of the neuropsychiatric diseases is mainly used for constructing a brain network based on a single connection relation (such as average anisotropy fraction among interested regions) to carry out comprehensive analysis so as to classify the neuropsychiatric diseases. The method can describe the connection relation among the brain network nodes from a certain angle, but due to the complexity of the human brain, the connection relation of each component in the brain can not be comprehensively expressed by a single connection relation (such as average anisotropy fraction among interested regions), so that the classification performance can be influenced.
Disclosure of Invention
In order to solve the technical problems, the application provides a neuropsychiatric disease classification method based on brain network analysis, which obtains original characteristic parameters by analyzing original data, then constructs a brain network according to the original characteristic parameters, and realizes the classification of neuropsychiatric diseases according to network structure characteristic parameters of the brain network.
The technical scheme adopted by the application is as follows: a neuropsychiatric disease classification method based on brain network analysis comprises the following steps:
s1, dividing the cerebral cortex and the sub-cortical structure into a plurality of interested regions through a brain atlas, and extracting connection characteristic parameters among the plurality of interested regions;
s2, constructing a multi-layer brain network according to the region of interest and the connection characteristic parameters;
s3, extracting network structure characteristic parameters in each layer of brain network;
the network structure characteristic parameters comprise: local clustering coefficients, node importance scores and node local importance scores;
s4, fusing the characteristics of the multilayer brain network based on the characteristic parameters of the network structure and by using the difference between the sample to be classified and the known classification sample compared with the normal sample to obtain the relative characteristic quantity of the multilayer brain network between the sample to be classified and the known classification sample; then calculating the distance from the sample to be classified to all known classification samples;
s5, selecting k known classification samples closest to the samples to be classified;
s6, if the k known classification samples belong to the same classification and the distance from the sample to be classified to the classification is smaller than the diameter of the classification, judging that the sample to be classified belongs to the classification, and adding the sample to be classified into the classification to which the known classification samples belong;
otherwise, judging whether the distance from the sample to be classified to k known classification samples in the classification is larger than the diameter of the classification, if so, establishing a new classification, and classifying the sample to be classified into the new classification; otherwise, ending.
Further, the connection characteristic parameters in step S1 include: the number of fiber bundles between each region of interest, the average length of the fiber bundles between each region of interest, and the average anisotropy fraction of voxels between each region of interest along the fiber bundle direction.
Further, in step S3, the local clustering coefficient calculation formula is:
wherein, p represents the serial number of each layer brain network; i represents a node serial number,p-th layer representing sample mThe total number of nodes adjacent to the node i in the brain network;a set of all nodes adjacent to node i in the p-th brain network representing sample m;representing the normalized edge weight value between the nodes i and j in the p-th brain network of the sample m;representing the normalized edge weight between the nodes i and k in the p-th brain network of the sample m;and (4) representing the normalized edge weight value between the nodes j and k in the p-th brain network of the sample m.
Further, in step S3, the node importance score is calculated as:
wherein,to representThe overall efficiency of the system is improved,a layer p brain network representing sample m;to representThe overall efficiency of the system is improved,the network of the p-th layer brain network representing the sample m is the network after the node i and the edge connected with the node i are removed; m represents the number of regions of interest;to representA set of all nodes in;to representRemoving the residual set after the node i;representing the shortest path length between the nodes s and t in the p-th brain network of the sample m;represents the shortest path length between nodes s ', t' in the p-th brain network of sample m.
Further, in step S3, the calculation formula of the node local importance score is:
wherein,to representThe overall efficiency of the system is improved,a local network of nodes i in a layer p brain network representing a sample m;to representThe overall efficiency of the system is improved,to representRemoving the node i and the network behind the edge connected with the node i;representing the number of nodes in a local network of the node i in the p-th brain network of the sample m;representing a set of all nodes in a local network of node i;to representRemoving the residual set after the node i;representing the shortest path length between the nodes s and t in the p-th brain network of the sample m;represents the shortest path length between nodes s ', t' in the p-th brain network of sample m.
Still further, the step S4 includes the following sub-steps:
s41, calculating the relative characteristic quantity of the multilayer network of the sample to be classified and the known classification sample according to the network structure characteristic parameter of each node of each layer of brain network of the sample to be classified;
s42, calculating cosine correlation between the multi-layer network relative characteristic quantities of the samples to be classified obtained in the step S41 and the known classification samples;
and S43, calculating the distance between the sample to be classified and the known classification sample according to the cosine correlation.
Further, the step S41 of calculating the relative feature quantity of the sample to be classified specifically includes:
b1, respectively calculating the validity weight of each type of network structure characteristic parameter of each node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in each layer of brain network according to the network structure characteristic parameter of each layer of brain network of the sample to be classified and the mean value and standard deviation of the network structure characteristic parameters of each layer of brain network of all normal samples:
when the validity weight of a certain type of network structure characteristic parameter of a certain node in a multilayer brain network corresponding to a sample to be classified or a known classification sample in a certain layer brain network is 1, indicating that the network structure characteristic parameter of the certain node in the multilayer brain network corresponding to the sample to be classified or the known classification sample is valid in the layer brain network;
when the validity weight of a certain type of network structure characteristic parameter of a certain node in a multilayer brain network corresponding to a sample to be classified or a known classification sample is 0, the network structure characteristic parameter of the certain node in the multilayer brain network corresponding to the sample to be classified or the known classification sample is invalid in the multilayer brain network;
b2, calculating the relative weight of each network structure characteristic parameter of each node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in each layer brain network:
if the network structure characteristic parameter of a certain node in the multilayer brain network corresponding to the sample to be classified and a known classification sample are invalid in each layer of brain network, the relative weight of the network structure characteristic parameter of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in each layer of brain network is 0;
if the network structure characteristic parameter of a certain node in the multilayer brain network corresponding to the sample to be classified and a known classification sample are effective in the certain layer brain network, the relative weight of the network structure characteristic parameter of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in the layer brain network is 1, and the relative weight of the network structure characteristic parameters in the rest layers brain networks is 0;
if the network structure characteristic parameter of a certain node in the multilayer brain network corresponding to one of the sample to be classified and a known classification sample is invalid in each layer of brain network, and the network structure characteristic parameter of a corresponding node in the other corresponding multilayer brain network is valid in each layer of brain network, the relative weight of the network structure characteristic parameter of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in the layer of brain network is 1, and the relative weight of the network structure characteristic parameter of the other layers of brain networks is 0;
if the network structure characteristic parameters of a certain node in the multilayer brain network corresponding to the sample to be classified and a known classification sample are respectively effective in different layers of brain networks, calculating the relative weight of the network structure characteristic parameters of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in the layer of brain network according to the effective coefficients corresponding to the layers of brain networks, wherein the relative weight of the other layers of brain networks is 0;
b3, calculating the relative characteristic quantity of the multi-layer network of the sample to be classified and the known classification sample according to the relative weight of the sample to be classified and the known classification sample, the effectiveness weight of each layer of brain network and the network structure characteristic parameter of the normal sample.
Further, in step B2, the relative weight between the sample to be classified and a known classification sample in the brain network of the layer is calculated according to the corresponding significant coefficients of each layer; the method specifically comprises the following steps:
setting effective coefficients corresponding to all layers of brain networks, setting a certain layer of brain network with the network structure characteristic parameter of the node effective in the multi-layer brain network corresponding to the sample to be classified, wherein the relative weight of the network structure characteristic parameter of the node in the multi-layer brain network corresponding to the sample to be classified and the known classification sample in the layer of brain network is as follows: dividing the effective coefficient corresponding to the layer brain network by the sum of the effective coefficient corresponding to the layer brain network and the effective coefficient corresponding to the layer brain network with the effective network structure characteristic parameters of the corresponding nodes in the known classification sample;
in the layer of brain network in which the network structure characteristic parameter of the node in the multilayer brain network corresponding to the known classification sample is effective, the relative weight of the sample to be classified and the network structure characteristic parameter of the node in the multilayer brain network corresponding to the known classification sample in the layer of brain network is as follows: and dividing the effective coefficient corresponding to the layer of brain network by the sum of the effective coefficient corresponding to the layer of brain network and the effective coefficient corresponding to the layer of brain network with the effective network structure characteristic parameter of the node in the sample to be classified.
Further, the initial diameter of the new classification established in step S6 is 0.
Further, after the step S6 of adding the sample to be classified into the classification to which the known classification sample belongs, the method further includes: the diameter of the classification is updated.
The invention has the beneficial effects that: the neural and mental disease classification method based on brain network analysis aims at the problem of inaccurate classification of the existing neural and mental disease classification, a brain atlas is divided into a plurality of regions of interest with specific functions, a plurality of layers of brain networks are obtained by respectively establishing brain networks reflecting the connection relation among the regions of interest, local features and global features of each layer of brain network are described by calculating network structure feature parameters, finally, fusion of the characteristics of the plurality of layers of brain networks is carried out by utilizing the difference between a sample to be classified and a known classification sample compared with a normal sample, and the distance among the samples is calculated to distinguish different types of neural and mental diseases; compared with the prior art, the method realizes that the connection relation among all the interested regions in the brain is described more comprehensively, not only can be used for correctly classifying the neuropsychiatric diseases with known pathology and definite characteristics, but also can be used for effectively classifying and identifying the neuropsychiatric diseases with unknown pathology and unclear characteristics or the unknown neuropsychiatric diseases with similar characteristics to the known neuropsychiatric diseases.
Drawings
FIG. 1 is a schematic flow chart of the present application;
FIG. 2 is an overall flow chart provided by an embodiment of the present application;
fig. 3 is a schematic diagram of a three-layer network model provided in an embodiment of the present application;
fig. 4 is a flowchart for calculating relative feature values according to an embodiment of the present application.
Detailed Description
In order to facilitate the understanding of the technical contents of the present invention by those skilled in the art, the present invention will be further explained with reference to the accompanying drawings.
The present application focuses primarily on the millimeter scale; after functional magnetic resonance imaging (fMRI) is performed on the brain and fMRI data is processed using professional medical image processing software and a specific brain Atlas (Atlas), the cerebral cortex and sub-cortical structures can be divided into a plurality of regions of interest, such as brain regions with specific functions. The regions of interest have clear anatomical structure and function, and different division modes of the regions of interest can be obtained by using different brain atlases. Neuropsychiatric disorders often cause changes in the nerves of the brain, such that the brain network may change in some relevant characteristic. It is generally believed that: different brain diseases have certain specificity. These specificities make their corresponding brain network characteristics different. The common brain diseases comprise mild cognitive impairment, Alzheimer's disease, children hyperactivity disorder, depression and the like.
As shown in fig. 1, a scheme flow chart of the present application is provided, and the technical scheme of the present application is as follows: a neuropsychiatric disease classification method based on brain network analysis comprises the following steps:
s1, dividing the cerebral cortex and the sub-cortical structure into a plurality of interested regions through a brain atlas, and extracting connection characteristic parameters among the plurality of interested regions;
s2, constructing a multi-layer brain network according to the region of interest and the connection characteristic parameters;
s3, extracting network structure characteristic parameters in each layer of brain network;
the network structure characteristic parameters comprise: local clustering coefficients, node importance scores and node local importance scores;
s4, fusing the characteristics of the multilayer brain network based on the characteristic parameters of the network structure and by using the difference between the sample to be classified and the known classification sample compared with the normal sample to obtain the relative characteristic quantity of the multilayer brain network between the sample to be classified and the known classification sample; then calculating the distance from the sample to be classified to all known classification samples;
s6, if the k known classification samples belong to the same classification and the distance from the sample to be classified to the classification is smaller than the diameter of the classification, judging that the sample to be classified belongs to the classification, and adding the sample to be classified into the classification to which the known classification samples belong;
otherwise, judging whether the distance from the sample to be classified to k known classification samples in the classification is larger than the diameter of the classification, if so, establishing a new classification, and classifying the sample to be classified into the new classification; otherwise, the further classification processing combined with the medical knowledge is required, but the further classification processing combined with the medical knowledge is not the focus of the application, and the execution of the method of the application is finished.
As shown in fig. 2, which is a detailed flowchart of the scheme of the present application, step S1 is to pre-process the sample to be classified, specifically: the method and the device have the advantages that functional magnetic resonance imaging (fMRI) data are obtained by performing functional magnetic resonance imaging on the brain, and characteristic parameters required by the scheme are obtained by analyzing through professional medical image processing software.
In the professional medical image processing software, a brain Atlas (Atlas) can be used for dividing cerebral cortex and sub-cortical structures into a plurality of interested regions, and different brain atlases correspondingly obtain different dividing modes of the interested regions. Common brain maps are: Harvard-Oxford clinical and subclinical clinical protocols atlas maps drawn by the centre of morphological analysis at Harvard university and JHU DTI-basedwhite-patient protocols proposed by the brain anatomical MRI laboratory at John Hopkins university, regions of interest segmented according to these maps have anatomically clear structures and functions.
The method uses three initial characteristic parameters to represent the connection relation between the interested regions from different angles, wherein the three initial characteristic parameters are respectively the number of the fiber bundles between the interested regions, the average length of the fiber bundles between the interested regions and the average anisotropic fraction of voxels between the interested regions along the fiber bundle direction, and the three parameters are called as connection characteristic parameters and are respectively written as: FA. FN and FL, which are the anisotropy fraction, number, and average length of the fiber bundle between the respective regions of interest, respectively.
FA (fractional anisotropy), an anisotropy fraction, is a mean anisotropy characterizing a voxel along the direction of the fiber bundle; the tight condition and the complete condition of the arrangement of the brain nerve fibers in the same direction are reflected, the value range of FA is 0-1, and the higher the value is, the tighter and more complete the nerve fibers in the corresponding area are.
Fn (fiber number), the number of nerve fibers, indicates the number of nerve fibers connecting two brain regions.
Fl (fiber length), the average nerve fiber length, represents the average length of nerve fibers connecting two different brain regions.
The three connection characteristic parameters represent the nerve conduction function strength among the interested regions from different angles, and the larger the values of FA, FN and FL are, the stronger the corresponding nerve conduction function among the interested regions is. The distribution rule of different neuropsychiatric diseases in FA, FL and FN has certain specificity.
Step S2 specifically includes: constructing a multi-layer brain network using the region of interest and the connection feature parameters; for each connection characteristic parameter, a layer of brain network reflecting the connection relation between the regions of interest from a specific angle can be constructed; using three connection feature parameters results in a multi-layer brain network with three levels. For each layer of the multi-layer brain network, three types of network structure characteristic parameters are selected to describe local characteristics and global characteristics of the brain network of the layer. The three types of network structure characteristic parameters selected by the application are respectively as follows: a local clustering coefficient, a node importance score, and a node local importance score.
In each layer of brain network, nodes in the network are all interested areas, edges in the network are connections among the interested areas, and the weight values of the edges are represented by characteristic parameters representing the strength of the connection relation among the interested areas. If the value of the connection characteristic parameter between the two interested areas is 0, no edge exists between the two interested areas; otherwise, an edge with the value of the connection characteristic parameter as a weight exists between the two interested areas.
Using three connection feature parameters results in a multi-layer brain network with three levels. The first, second and third layers use FA, FN and FL as connection characteristic parameters.
In each layer of the multilayer network, all nodes correspond to the same group of interest one by one, the number of interest obtained in the present application is denoted as M, the number of nodes in each layer is also denoted as M, and the nodes are numbered as i (i is 1,2, …, M). The multi-layer brain network model is shown in fig. 3.
In FIG. 3, useRepresents the edge weight between nodes i and j in the p-th brain network of the sample m (when the nodes i and j are not communicated, i.e. there is no edge between the nodes i and j, let). The edge weight value between two nodes in the p-th brain network is the value of the connection characteristic parameter corresponding to the two nodes in the layer. The following connection characteristic parameter matrix can be used to represent the edge weights between nodes in each layer of brain network.
Let p-th layer brain network (p ═ 1,2,3) beNote the bookThe set of all nodes in the node ism represents the number of samples (samples to be classified).
Step S3 specifically includes: extraction of network structure characteristic parameters
According to the method, three types of network structure characteristic parameters, namely a local clustering coefficient, a node importance score and a node local importance score, are selected to describe local characteristics and global characteristics in each layer of brain network. Local clustering coefficients and node local importance scores are adopted to describe local features in each layer of brain network, and node importance scores are adopted to describe global features in each layer of brain network.
For each layer of the multi-layer brain network, M characteristic parameter values can be calculated by each type of network structure characteristic parameters, and 3 × M characteristic parameter values are generated by 3 types of network structure characteristic parameters. The following describes in detail the extraction process of each network structure characteristic parameter:
1) local clustering coefficient
In the brain network, the clustering coefficient of a certain node, i.e. the clustering coefficient of a region of interest, represents the degree of closeness of the connection with the surrounding region of interest. The higher the clustering coefficient value of a node, the more closely the node is associated with its neighboring nodes.
Note that the set of all nodes adjacent to node i in the sample mth layer brain network is(p＝1,2,3)。
In the p-th layer brain network of the sample m, the local clustering coefficient of the ith node in the p (p is 1,2,3) th layer brain networkThe calculation formula is as follows:
wherein,
in the formula (1), the first and second groups,representing the total number of nodes adjacent to the node i in the p-th brain network of the sample m;the value range of (1) is (0),representing the normalized edge weight value between the node i and the node j in the p-th brain network of the sample m;representing the normalized edge weight between the node i and the node k in the p-th brain network of the sample m;and representing the normalized edge weight value between the node j and the node k in the p-th brain network of the sample m.
In the formula (2), the first and second groups,representing the set of all nodes in the p-th brain network of sample m,representing the edge weight value between the nodes i and j in the p-th brain network of the sample m;the maximum value of the inter-node edge weights in the sample mth layer brain network is s, t are two different nodes in the sample mth layer brain network;and the normalized edge weight value between the nodes i and j in the p-th brain network of the sample m.
2) Node importance score
Will be provided withA network in which node i and edges between node i and other nodes are removed from the network is referred to asThe set of points with i removed is denoted asNamely, it is
Defining a parameter for each node i of the sample m layer p brain network: node importance scores, recorded asThe node importance scores may be used to quantitatively measure the magnitude of the importance of each node in the network. In brain networks, the node importance score reflects the degree of importance of a certain region of interest in the entire brain network. The calculation method is as follows:
in the formula,andrespectively representing networksAndthe global efficiency reflects the average efficiency of mutual communication among nodes in the network, and M is the number of the interested areas obtained in the step one;representing the sum of the edge weights corresponding to the edges through which the shortest path between the nodes s and t passes in the p-th brain network of the sample m, namely the length of the shortest path between the nodes s and t;and the shortest path length between the nodes s 'and t' is represented as the sum of the edge weights corresponding to the edges passed by the shortest path between the nodes s 'and t' in the p-th brain network of the sample m.
Defining the shortest path between nodes s and t in the p-th brain network of the sample m as a path with the minimum sum of the edge weights corresponding to the edges passing through all the paths between the nodes s and t; the Dijkstra algorithm can be used to calculate the shortest path from node S to node t, let set S represent the set of nodes for which the shortest path has been found (initially, S has only node S, at this time, the set shortest path is from node S to node S, the shortest path length is 0), let U represent the set of vertices for which the shortest path is not determined,each node in the sets S and U corresponds to a distance, and the distance of the node in S is the shortest path length from S to the node. The distance of the node in U is the length of the current shortest path from S to the node, wherein the node only comprises the node in S as the intermediate node. If a node in U is not directly connected with all nodes in the current S set, the distance of the node in the current U is infinite. The algorithm comprises the following specific steps:
1. initially, S only includes a node S, i.e., S ═ S }, and the distance of the node S, i.e., the shortest path length from S to S (is 0). U comprisesExcept for s, i.e., U ═ the rest nodes. The distance dist (U) of a node U in U is defined as follows: if a node U in U is not directly connected to a node s, the distance between the node U and the node s in U is positive infinity (i.e., dist (U) ∞). If a certain node U in U is directly connected with a node s (namely, an edge exists between two nodes), the distance of the node in the U is the edge weight between the node s and the node U, namely
2. Selecting a node v with the minimum distance from s from U, and requiringV is removed from U and added to S (the distance of the newly added node v is the shortest path length from v to S).
3. And updating the distances of all nodes in the current set U: and if a certain node in the current U is not directly connected with the newly added node v, the distance of the node in the current U is unchanged. If a certain node U in U is directly connected with a newly added node v, comparing the distance of the node v with the sum of the edge weights between the node U and the node v in U (namely, comparing the distance between the node v and the node v in U with each other) And (4) taking the smaller value of the distance between the current U and the node U as the updated value of the distance between the current U and the node U.
4. And repeating the steps 2 and 3 until the shortest path from the node s to the node t is obtained.
Similarly, the shortest path between the nodes s 'and t' can be obtained according to the steps 1 to 4.
3) Node local importance score
First, a local network with respect to a node i is defined as follows: p-th layer brain network for sample mAny one of the nodes i, local network with respect to the node iThe system is composed of the following nodes and edges: the nodes forming the network are a node i and nodes adjacent to the node i; the edge forming the network is the original network of the nodesThe corresponding edge in (1). Note the set of all nodes in the local network for node iAre synthesized into(p ═ 1,2, 3); will be provided withA network in which node i and edges between node i and other nodes are removed from the network is referred to asThe set of points with i removed is denoted asNamely, it is
The node local importance score reflects how important a region of interest is in the network it and the adjacent regions of interest. For the ith node of the sample m p layer brain network, its node local importance score is recorded asThe calculation formula is as follows:
in the formula (4), the first and second groups,andrespectively representing networksAndthe overall efficiency of the system is improved,representing the number of nodes in a local network related to i in the p-th layer network of the sample m;represents the sum of the edge weights corresponding to the edges passed by the shortest path between the nodes s and t in the p-th brain network of the sample m, namely the length of the shortest path between the nodes s and t,and the shortest path length between the nodes s 'and t' is represented as the sum of the edge weights corresponding to the edges passed by the shortest path between the nodes s 'and t' in the p-th brain network of the sample m.
By calculating the three types of network structure characteristic parameters, each layer of network can obtain 3 × M network structure characteristic parameters, and each multilayer brain network can obtain 9 × M characteristic parameters.
Step S4 specifically includes: fusing multilayer network characteristics and calculating the distance between a sample to be classified and a known classification sample;
in order to facilitate subsequent expression, three types of network structure characteristic parameters are numbered by using a variable q, wherein q is 1,2 and 3 respectively represent a local clustering coefficient, a node importance score and a node local importance score. The q-th network structure characteristic parameters of the sample m p-th network node i are recorded asFor normal sample set RnorThe q-th network structure characteristic parameters of the p-th network node i are calculated, and the average value of the q-th network structure characteristic parameters is calculatedAnd standard deviation ofThe calculation method is as follows:
the distance between two samples, i.e. the distance between two multi-layer brain networks, is defined.
The method comprises the steps of firstly, comprehensively analyzing a normal sample and an abnormal sample, and calculating the relative characteristic quantity of the multilayer brain network of the two samples according to a plurality of network structure characteristic parameters of each layer of network of the sample. Based on the relative feature quantity of the multi-layer brain network of the two samples, the distance between the two samples is defined by combining cosine correlation. All samples to be classified are from patients confirmed by clinical diagnosis of a doctor, and are abnormal; after being processed by the method, a certain sample to be classified becomes a known classification sample, but the known classification sample is still an abnormal sample; namely, what is included in the known classification samples is the classified abnormal sample.
As shown in fig. 4, the specific steps are as follows:
firstly, carrying out data normalization processing on all samples according to a formula (7), and recording q-th network structure characteristic parameters in a p-th network node i of a normalized sample mThe calculation formula is as follows:
in the following, m and n represent the numbers of the two abnormal samples with the distance to be calculated respectively; all the classified or to-be-classified abnormal samples are patient data confirmed by clinical diagnosis of a doctor,(p ═ 1,2,3) is the set of all nodes in sample m layer p brain network;and the q-th network structure characteristic parameters represent a set formed by all nodes of the p-th network of the sample m.
Respectively recording relative characteristic quantities of the q-th class network structure characteristic parameters of the ith node in the sample m and the sample n multilayer brain networks asAnd
in the formula,andand (3) layer effectiveness weighting of a characteristic parameter of a class i and q network structure of a node in the sample multilayer brain network. Taking a sample m as an example, the method represents the effectiveness of the structural characteristic parameters of the class q network of the node i in the multilayer brain network of the sample m in the layer p, and the value taking method for each layer network is as follows:
wherein ^ is the combination of condition 1 and condition 2, i.e. condition 1 and condition 2;
andfor relative weights, the calculation method is as follows (four cases in total):
1) if p is present0So thatThen
2) If p is present1≠p2So thatThen
And is
In the formula, the relative weights of n to m and m to n of the p1 layers are the same, and the relative weights of n to m and m to n of the p2 layers are the same; k is a radical of1、k2、k3And respectively taking default values of effective coefficients corresponding to each layer of brain network: 1. 0.6, 0.3; k is a radical of1、k2、k3When p is 1,2,3 or p11,2,3 or p2When 1,2,3, kp、Is k1、k2、k3。
3) If it isAndof which only one has a value of 0 when p is 1,2,3 (assuming that it is) And there is p ═ p0So thatThen
4) If it isAndwhen p is 1,2,3, all values are 0, i.e. Then the process of the first step is carried out,
through the calculation, the relative characteristic quantity of the multilayer brain network of the two samples can be obtained. Each sample had a total of 3 × M relative features.
Then, calculating cosine correlation coefficient between q-th class network structure characteristic parameters of the multi-layer brain network corresponding to the two samples m and n according to the following formulaThe calculation method is as follows:
after cosine correlation coefficients between relative characteristic quantities of the same-type network structure characteristic parameters of the two samples are calculated, 3 cosine correlation coefficients representing 3 different network structure parameters between the two samples can be obtained, and then the 3 cosine correlation coefficients are utilized to calculate the distance between the two samples:
distance d between samples m, nm,nPositive values greater than 0.
Step S5 specifically includes: selecting k known classification samples (the default k is 5) nearest to the samples to be classified;
step S6 specifically includes: if the samples belong to a known classification and the distance from the sample to be classified to the classification is smaller than the diameter of the classification, judging that the sample belongs to the classification; if the distances from the sample to be classified to the k known classification samples are all larger than the diameters of the corresponding classifications (no matter whether the k known classification samples belong to the same known classification or not), judging that the sample belongs to a new classification, and establishing the new classification; otherwise, further processing is carried out by combining the knowledge in the medical aspect.
In the step, the diameter of a certain known classification Z is defined as the maximum value of the distance between every two samples in the classification Z; the distance from a sample to be classified to a certain class is defined as the average distance from the sample to be classified to all samples in the class.
If the sample to be classified is classified to a known classification, adding the sample to be classified into the currently classified set, and updating the diameter of the classification; if a new classification is established in step 3, the initial diameter of the new classification is 0.
It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to assist the reader in understanding the principles of the invention and are to be construed as being without limitation to such specifically recited embodiments and examples. Various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims (10)
1. A neuropsychiatric disease classification method based on brain network analysis is characterized by comprising the following steps:
s1, dividing the cerebral cortex and the sub-cortical structure into a plurality of interested regions through a brain atlas, and extracting connection characteristic parameters among the plurality of interested regions;
s2, constructing a multi-layer brain network according to the region of interest and the connection characteristic parameters;
s3, extracting network structure characteristic parameters in each layer of brain network;
the network structure characteristic parameters comprise: local clustering coefficients, node importance scores and node local importance scores;
s4, fusing the characteristics of the multilayer brain network based on the characteristic parameters of the network structure and by using the difference between the sample to be classified and the known classification sample compared with the normal sample to obtain the relative characteristic quantity of the multilayer brain network between the sample to be classified and the known classification sample; then calculating the distance from the sample to be classified to all known classification samples;
s5, selecting k known classification samples closest to the samples to be classified;
s6, if the k known classification samples belong to the same classification and the distance from the sample to be classified to the classification is smaller than the diameter of the classification, judging that the sample to be classified belongs to the classification, and adding the sample to be classified into the classification to which the known classification samples belong;
otherwise, judging whether the distance from the sample to be classified to k known classification samples in the classification is larger than the diameter of the classification, if so, establishing a new classification, and classifying the sample to be classified into the new classification; otherwise, ending.
2. The method for classifying neuropsychiatric diseases based on brain network analysis according to claim 1, wherein the connection feature parameters of step S1 include: the number of fiber bundles between each region of interest, the average length of the fiber bundles between each region of interest, and the average anisotropy fraction of voxels between each region of interest along the fiber bundle direction.
3. The method for classifying neuropsychiatric diseases based on brain network analysis according to claim 2, wherein the local clustering coefficient calculation formula in step S3 is:
<mrow> <msubsup> <mi>C</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <msubsup> <mi>k</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <msubsup> <mi>k</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mo>(</mo> <mi>i</mi> <mo>)</mo> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> </mfrac> <munder> <munder> <mo>&Sigma;</mo> <mrow> <mi>j</mi> <mo>,</mo> <mi>k</mi> <mo>&Element;</mo> <msubsup> <mi>V</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> </mrow> </mrow> </munder> <mrow> <mi>j</mi> <mo>&NotEqual;</mo> <mi>k</mi> </mrow> </munder> <msup> <mrow> <mo>(</mo> <msubsup> <mover> <mi>w</mi> <mo>^</mo> </mover> <mi>m</mi> <mi>p</mi> </msubsup> <mo>(</mo> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> <mo>)</mo> <msubsup> <mover> <mi>w</mi> <mo>^</mo> </mover> <mi>m</mi> <mi>p</mi> </msubsup> <mo>(</mo> <mrow> <mi>i</mi> <mo>,</mo> <mi>k</mi> </mrow> <mo>)</mo> <msubsup> <mover> <mi>w</mi> <mo>^</mo> </mover> <mi>m</mi> <mi>p</mi> </msubsup> <mo>(</mo> <mrow> <mi>j</mi> <mo>,</mo> <mi>k</mi> </mrow> <mo>)</mo> <mo>)</mo> </mrow> <mrow> <mn>1</mn> <mo>/</mo> <mn>3</mn> </mrow> </msup> </mrow>
wherein, p represents the serial number of each layer brain network;i represents a node serial number,representing the total number of nodes adjacent to the node i in the p-th brain network of the sample m;a set of all nodes adjacent to node i in the p-th brain network representing sample m;representing the normalized edge weight value between the nodes i and j in the p-th brain network of the sample m;representing the normalized edge weight between the nodes i and k in the p-th brain network of the sample m;and (4) representing the normalized edge weight value between the nodes j and k in the p-th brain network of the sample m.
4. The method for classifying neuropsychiatric diseases based on brain network analysis according to claim 3, wherein the node importance score is calculated in step S3 by:
<mrow> <msubsup> <mi>NIS</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mo>=</mo> <mfrac> <mrow> <msubsup> <mi>NETE</mi> <mi>m</mi> <mi>p</mi> </msubsup> </mrow> <mrow> <msubsup> <mi>NETE</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mfrac> <mn>1</mn> <mrow> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <mi>M</mi> <mrow> <mo>(</mo> <mi>M</mi> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> </mfrac> <munder> <munder> <mo>&Sigma;</mo> <mrow> <mi>s</mi> <mo>,</mo> <mi>t</mi> <mo>&Element;</mo> <msubsup> <mi>V</mi> <mi>m</mi> <mi>p</mi> </msubsup> </mrow> </munder> <mrow> <mi>s</mi> <mo>&NotEqual;</mo> <mi>t</mi> </mrow> </munder> <mfrac> <mn>1</mn> <mrow> <msubsup> <mi>d</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>s</mi> <mo>,</mo> <mi>t</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </mrow> <mrow> <mfrac> <mn>1</mn> <mrow> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <mrow> <mo>(</mo> <mi>M</mi> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mi>M</mi> <mo>-</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow> </mfrac> <munder> <munder> <mo>&Sigma;</mo> <mrow> <msup> <mi>s</mi> <mo>&prime;</mo> </msup> <mo>,</mo> <msup> <mi>t</mi> <mo>&prime;</mo> </msup> <mo>&Element;</mo> <msubsup> <mover> <mi>V</mi> <mo>~</mo> </mover> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> </mrow> </mrow> </munder> <mrow> <msup> <mi>s</mi> <mo>&prime;</mo> </msup> <mo>&NotEqual;</mo> <msup> <mi>t</mi> <mo>&prime;</mo> </msup> </mrow> </munder> <mfrac> <mn>1</mn> <mrow> <msubsup> <mi>d</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <msup> <mi>s</mi> <mo>&prime;</mo> </msup> <mo>,</mo> <msup> <mi>t</mi> <mo>&prime;</mo> </msup> <mo>)</mo> </mrow> </mrow> </mfrac> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mo>(</mo> <mi>M</mi> <mo>-</mo> <mn>2</mn> <mo>)</mo> <munder> <munder> <mo>&Sigma;</mo> <mrow> <mi>s</mi> <mo>,</mo> <mi>t</mi> <mo>&Element;</mo> <msubsup> <mi>V</mi> <mi>m</mi> <mi>p</mi> </msubsup> </mrow> </munder> <mrow> <mi>s</mi> <mo>&NotEqual;</mo> <mi>t</mi> </mrow> </munder> <mfrac> <mn>1</mn> <mrow> <msubsup> <mi>d</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>s</mi> <mo>,</mo> <mi>t</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </mrow> <mrow> <mi>M</mi> <munder> <munder> <mo>&Sigma;</mo> <mrow> <msup> <mi>s</mi> <mo>&prime;</mo> </msup> <mo>,</mo> <msup> <mi>t</mi> <mo>&prime;</mo> </msup> <mo>&Element;</mo> <msubsup> <mover> <mi>V</mi> <mo>~</mo> </mover> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> </mrow> </mrow> </munder> <mrow> <msup> <mi>s</mi> <mo>&prime;</mo> </msup> <mo>&NotEqual;</mo> <msup> <mi>t</mi> <mo>&prime;</mo> </msup> </mrow> </munder> <mfrac> <mn>1</mn> <mrow> <msubsup> <mi>d</mi> <mi>m</mi> <mi>p</mi> </msubsup> <mrow> <mo>(</mo> <msup> <mi>s</mi> <mo>&prime;</mo> </msup> <mo>,</mo> <msup> <mi>t</mi> <mo>&prime;</mo> </msup> <mo>)</mo> </mrow> </mrow> </mfrac> </mrow> </mfrac> </mrow>
wherein,to representGlobal effect ofThe ratio of the total weight of the particles,a layer p brain network representing sample m;to representThe overall efficiency of the system is improved,the network of the p-th layer brain network representing the sample m is the network after the node i and the edge connected with the node i are removed; m represents the number of regions of interest;to representA set of all nodes in;to representRemoving the residual set after the node i;representing the shortest path length between the nodes s and t in the p-th brain network of the sample m;represents the shortest path length between nodes s ', t' in the p-th brain network of sample m.
5. The method for classifying neuropsychiatric diseases based on brain network analysis according to claim 4, wherein the calculation formula of the node local importance score in step S3 is:
wherein,to representThe overall efficiency of the system is improved,a local network of nodes i in a layer p brain network representing a sample m;to representThe overall efficiency of the system is improved,to representRemoving the node i and the network behind the edge connected with the node i;representing the number of nodes in a local network of the node i in the p-th brain network of the sample m;representing a set of all nodes in a local network of node i;to representRemoving the residual set after the node i;representing the shortest path length between the nodes s and t in the p-th brain network of the sample m;represents the shortest path length between nodes s ', t' in the p-th brain network of sample m.
6. The method for classifying neuropsychiatric diseases based on brain network analysis according to claim 1, wherein the step S4 comprises the following substeps:
s41, calculating the relative characteristic quantity of the multilayer network of the sample to be classified and the known classification sample according to the network structure characteristic parameter of each node of each layer of brain network of the sample to be classified;
s42, calculating cosine correlation between the multi-layer network relative characteristic quantities of the samples to be classified obtained in the step S41 and the known classification samples;
and S43, calculating the distance between the sample to be classified and the known classification sample according to the cosine correlation.
7. The method for classifying neuropsychiatric diseases based on brain network analysis according to claim 6, wherein the step S41 is specifically configured to calculate the relative feature quantities of the samples to be classified as follows:
b1, respectively calculating the validity weight of each type of network structure characteristic parameter of each node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in each layer of brain network according to the network structure characteristic parameter of each layer of brain network of the sample to be classified and the mean value and standard deviation of the network structure characteristic parameters of each layer of brain network of all normal samples:
when the validity weight of a certain type of network structure characteristic parameter of a certain node in a multilayer brain network corresponding to a sample to be classified or a known classification sample in a certain layer brain network is 1, indicating that the network structure characteristic parameter of the certain node in the multilayer brain network corresponding to the sample to be classified or the known classification sample is valid in the layer brain network;
when the validity weight of a certain type of network structure characteristic parameter of a certain node in a multilayer brain network corresponding to a sample to be classified or a known classification sample is 0, the network structure characteristic parameter of the certain node in the multilayer brain network corresponding to the sample to be classified or the known classification sample is invalid in the multilayer brain network;
b2, calculating the relative weight of each network structure characteristic parameter of each node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in each layer brain network:
if the network structure characteristic parameter of a certain node in the multilayer brain network corresponding to the sample to be classified and a known classification sample are invalid in each layer of brain network, the relative weight of the network structure characteristic parameter of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in each layer of brain network is 0;
if the network structure characteristic parameter of a certain node in the multilayer brain network corresponding to the sample to be classified and a known classification sample are effective in the certain layer brain network, the relative weight of the network structure characteristic parameter of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in the layer brain network is 1, and the relative weight of the network structure characteristic parameters in the rest layers brain networks is 0;
if the network structure characteristic parameter of a certain node in the multilayer brain network corresponding to one of the sample to be classified and a known classification sample is invalid in each layer of brain network, and the network structure characteristic parameter of a corresponding node in the other corresponding multilayer brain network is valid in each layer of brain network, the relative weight of the network structure characteristic parameter of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in the layer of brain network is 1, and the relative weight of the network structure characteristic parameter of the other layers of brain networks is 0;
if the network structure characteristic parameters of a certain node in the multilayer brain network corresponding to the sample to be classified and a known classification sample are respectively effective in different layers of brain networks, calculating the relative weight of the network structure characteristic parameters of the node in the multilayer brain network corresponding to the sample to be classified and the known classification sample in the layer of brain network according to the effective coefficients corresponding to the layers of brain networks, wherein the relative weight of the other layers of brain networks is 0;
b3, calculating the relative characteristic quantity of the multi-layer network of the sample to be classified and the known classification sample according to the relative weight of the sample to be classified and the known classification sample, the effectiveness weight of each layer of brain network and the network structure characteristic parameter of the normal sample.
8. The method according to claim 7, wherein the step B2 is to calculate the relative weight between the sample to be classified and a known classification sample in each layer of brain network according to the significant coefficients corresponding to the brain networks; the method specifically comprises the following steps:
setting effective coefficients corresponding to all layers of brain networks, setting a certain layer of brain network with the network structure characteristic parameter of the node effective in the multi-layer brain network corresponding to the sample to be classified, wherein the relative weight of the network structure characteristic parameter of the node in the multi-layer brain network corresponding to the sample to be classified and the known classification sample in the layer of brain network is as follows: dividing the effective coefficient corresponding to the layer brain network by the sum of the effective coefficient corresponding to the layer brain network and the effective coefficient corresponding to the layer brain network with the effective network structure characteristic parameters of the corresponding nodes in the known classification sample;
in the layer of brain network in which the network structure characteristic parameter of the node in the multilayer brain network corresponding to the known classification sample is effective, the relative weight of the sample to be classified and the network structure characteristic parameter of the node in the multilayer brain network corresponding to the known classification sample in the layer of brain network is as follows: dividing the effective coefficient corresponding to the layer of brain network by the sum of the effective coefficient corresponding to the layer of brain network and the effective coefficient corresponding to the layer of brain network with the effective network structure characteristic parameters of the corresponding nodes in the sample to be classified;
the relative weight of the class network structure characteristic parameter of the node in the multi-layer brain network corresponding to the classification sample and the known classification sample in the rest layers of brain networks is 0.
9. The method for classifying neuropsychiatric disease based on brain network analysis according to claim 1, wherein the initial diameter of the new classification established in step S6 is 0.
10. The method for classifying neuropsychiatric diseases based on brain network analysis according to claim 9, wherein the step S6 further comprises, after adding the sample to be classified into the classification to which the known classification sample belongs: the diameter of the classification is updated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710940330.2A CN107680677B (en) | 2017-10-11 | 2017-10-11 | Neuropsychiatric disease classification method based on brain network analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710940330.2A CN107680677B (en) | 2017-10-11 | 2017-10-11 | Neuropsychiatric disease classification method based on brain network analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107680677A true CN107680677A (en) | 2018-02-09 |
| CN107680677B CN107680677B (en) | 2020-09-15 |
Family
ID=61140083
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710940330.2A Active CN107680677B (en) | 2017-10-11 | 2017-10-11 | Neuropsychiatric disease classification method based on brain network analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107680677B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110880008A (en) * | 2018-09-06 | 2020-03-13 | 刘艳 | Structural feature extraction and classification method based on brain nuclear magnetic image data |
| CN110916615A (en) * | 2019-12-05 | 2020-03-27 | 四川大学 | Alzheimer's disease detection device and computer equipment |
| CN112102952A (en) * | 2020-08-24 | 2020-12-18 | 广州金域医学检验中心有限公司 | Method for identifying pathological category based on distance calculation method and related device |
| CN113344883A (en) * | 2021-06-10 | 2021-09-03 | 华南师范大学 | Multilayer morphological brain network construction method, intelligent terminal and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105163659A (en) * | 2013-05-01 | 2015-12-16 | 株式会社国际电气通信基础技术研究所 | Brain activity analysis device, brain activity analysis method, and biomarker device |
| CN105726026A (en) * | 2016-01-28 | 2016-07-06 | 电子科技大学 | Mild cognitive impairment disease classifying method based on brain network and brain structure information |
| CN106295709A (en) * | 2016-08-18 | 2017-01-04 | 太原理工大学 | Functional magnetic resonance imaging data classification method based on multiple dimensioned brain network characterization |
| CN106845137A (en) * | 2017-02-27 | 2017-06-13 | 浙江工业大学 | A kind of sacred disease analysis method based on brain network structure |
| WO2017149542A1 (en) * | 2016-03-01 | 2017-09-08 | Sentimetrix, Inc | Neuropsychological evaluation screening system |
-
2017
- 2017-10-11 CN CN201710940330.2A patent/CN107680677B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105163659A (en) * | 2013-05-01 | 2015-12-16 | 株式会社国际电气通信基础技术研究所 | Brain activity analysis device, brain activity analysis method, and biomarker device |
| CN105726026A (en) * | 2016-01-28 | 2016-07-06 | 电子科技大学 | Mild cognitive impairment disease classifying method based on brain network and brain structure information |
| WO2017149542A1 (en) * | 2016-03-01 | 2017-09-08 | Sentimetrix, Inc | Neuropsychological evaluation screening system |
| CN106295709A (en) * | 2016-08-18 | 2017-01-04 | 太原理工大学 | Functional magnetic resonance imaging data classification method based on multiple dimensioned brain network characterization |
| CN106845137A (en) * | 2017-02-27 | 2017-06-13 | 浙江工业大学 | A kind of sacred disease analysis method based on brain network structure |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110880008A (en) * | 2018-09-06 | 2020-03-13 | 刘艳 | Structural feature extraction and classification method based on brain nuclear magnetic image data |
| CN110880008B (en) * | 2018-09-06 | 2023-10-17 | 刘艳 | Structural feature extraction and classification method based on brain nuclear magnetic image data |
| CN110916615A (en) * | 2019-12-05 | 2020-03-27 | 四川大学 | Alzheimer's disease detection device and computer equipment |
| CN112102952A (en) * | 2020-08-24 | 2020-12-18 | 广州金域医学检验中心有限公司 | Method for identifying pathological category based on distance calculation method and related device |
| CN112102952B (en) * | 2020-08-24 | 2024-05-14 | 广州金域医学检验中心有限公司 | Method for identifying pathology category based on distance calculation method and related equipment |
| CN113344883A (en) * | 2021-06-10 | 2021-09-03 | 华南师范大学 | Multilayer morphological brain network construction method, intelligent terminal and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107680677B (en) | 2020-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wen et al. | MVS-GCN: A prior brain structure learning-guided multi-view graph convolution network for autism spectrum disorder diagnosis | |
| Islam et al. | Brain MRI analysis for Alzheimer’s disease diagnosis using an ensemble system of deep convolutional neural networks | |
| CN109376751B (en) | Human brain function network classification method based on convolutional neural network | |
| Khojasteh et al. | Fundus images analysis using deep features for detection of exudates, hemorrhages and microaneurysms | |
| Mohammed et al. | Review on Nasopharyngeal Carcinoma: Concepts, methods of analysis, segmentation, classification, prediction and impact: A review of the research literature | |
| Li et al. | Pooling regularized graph neural network for fmri biomarker analysis | |
| Pouliakis et al. | Artificial neural networks as decision support tools in cytopathology: past, present, and future | |
| Marzullo et al. | Classification of multiple sclerosis clinical profiles via graph convolutional neural networks | |
| CN107680677B (en) | Neuropsychiatric disease classification method based on brain network analysis | |
| Hajamohideen et al. | Four-way classification of Alzheimer’s disease using deep Siamese convolutional neural network with triplet-loss function | |
| Couvy-Duchesne et al. | Ensemble learning of convolutional neural network, support vector machine, and best linear unbiased predictor for brain age prediction: Aramis contribution to the predictive analytics competition 2019 challenge | |
| Kiliçarslan et al. | Detection and classification of pneumonia using novel Superior Exponential (SupEx) activation function in convolutional neural networks | |
| Yang et al. | Large-scale brain functional network integration for discrimination of autism using a 3-D deep learning model | |
| Aslam et al. | Neurological Disorder Detection Using OCT Scan Image of Eye | |
| Liu et al. | DeepBundle: fiber bundle parcellation with graph convolution neural networks | |
| Lan et al. | Multi-view convolutional neural network with leader and long-tail particle swarm optimizer for enhancing heart disease and breast cancer detection | |
| Lella et al. | Communicability distance reveals hidden patterns of Alzheimer’s disease | |
| Xi et al. | The effect of high-order interactions on the functional brain networks of boys with ADHD | |
| Heitmann et al. | DeepBreath—automated detection of respiratory pathology from lung auscultation in 572 pediatric outpatients across 5 countries | |
| Osmanlıoğlu et al. | Connectomic assessment of injury burden and longitudinal structural network alterations in moderate‐to‐severe traumatic brain injury | |
| Xi et al. | Brain Functional Networks with Dynamic Hypergraph Manifold Regularization for Classification of End-Stage Renal Disease Associated with Mild Cognitive Impairment. | |
| Lamani et al. | Multi-atlas graph convolutional networks and convolutional recurrent neural networks-based ensemble learning for classification of autism spectrum disorders | |
| Ouchicha et al. | A novel deep convolutional neural network model for Alzheimer’s disease classification using brain MRI | |
| Mouches et al. | Multimodal brain age prediction fusing morphometric and imaging data and association with cardiovascular risk factors | |
| Ganeshkumar et al. | Two-stage deep learning model for automate detection and classification of lung diseases |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |